There are a couple of reasons I can think of that may cause this issue.
Important is that there are no RP-SMA/SMA conversion issues. Also, uFL might be a trouble maker.
My question is: how is your terrain? I would expect that at a distance of 10 meters of your antenna you get RSSI somewhere in the range of -30 to -60 dBm. is this true?
Also, can you verify your LGT92 antenna?
All connection tight (use tools not fingers!). Such high gain ants can show notching…have you tested at a distance? Have you looked at polar plots? Some have up tilt some down, some flat…worst case is one with a notched coverage pattern and uptilt… in this situation I start with substituting a simple ant closer to isotropic coverage say a 2 or 3dbi ant off a bit of kit (in right frequency band of course…you did check tuning/band on main ant right?)
Update: Also just looked at pic… the lgt is designed to hang from a lanyard… remove from close proximity metalwork which may be detuning and hang vertically form a nail in a wooden post say 1.5-2m above ground then check rssi’s again looking for values similar to @pe1mew ‘s shout. Personally I would expect -50 - -65 around there… lgt92 ant isn’t great but have had many 5-15km hits & and a few on recent vac with great LoS in the 30-50km range! But can be easily defined…
To add to @Jeff-UK response. Check the SMA connectors have the correct pin and sockets in the centre. It’s so easy to have two SMA connectors both with sockets and therefore a poor RF connection. (Even did this myself!! What was worse, I was the presenter at the antenna workshop 
)
Nope, I can go for miles and miles and miles.
All the base chips are from the same source, ie Semtech.
RAK have been building concentrators for their extensive range for quite a while now so it’s not the product design and it is unlikely to be faulty but we will find out along the way.
My extras:
The position of the antenna in the LGT92 may not be optimal when parked on a big metal tube - as mentioned above, try it hanging on it’s lanyard.
Use the gateway as it would be out the box - just it and the antenna it came with on the uFL lead. This should be a very quick test to do. May be simpler if it’s parked on the window sill of a ground floor room, bearing in mind the best reception will be on that side of the house.
If the range is still poor, please be aware that uFL was designed by an evil committee of software engineers - ideally the right tool, it not some slim needle nose pliers, to take off the connector and check it’s inner connector is still OK.
To echo Nick’s comment, and fact I see you also have a dgno lht…take that ant and add straight to the ufl to sma (with form marching adapter if needed taking into account pinned vs female comments!) and as Nick says put on windowsill with ant vertical… or if you trust weather forecast for quick field test outside (plastic bag just in case?!) 
Not using RP-SMA. I bought a custom cable from Antenna Web and have a normal male SMA on the end, then a female SMA to UFL adapter.
Terrain wise, i’m in suburban Kansas City. We’ve got some big mature trees, but not super dense. The location I lost signal is actually uphill from my house slightly, but if I were to get above the trees, I could see my house and antenna. So it’s not LOS perfect, but i’m not trying to get signal behind a mountain or through a thick forest or anything.
I did some testing with the LHT65. The LGT92 was not working right for me, I think it’s batt was going dead. But this is what I came up with from the antenna: 
Just a basic test around the house and even when I took the antenna/Pi outside with the stock antenna, I used the Dragino, LOS, about 20 ft away, level ground, and was getting -70 rssi with the LGT92. Again, stock antenna, outside on the ground (in a box but the antenna was sticking up) 20 ft away.
So i’m kinda stumped. It definitely seems to do better in the NE direction (the -63) but why not all the others? It’s not a mono-directional antenna…
Another interesting development! I took my TTN indoor gateway, took it apart and hooked up my antenna setup to it’s antenna UFL connection. I took my LHT65 out and I was getting 7-9 db better signal in the same test spots. I was getting consistently 67 db on the opposite side of the cul-de-sac. I got my LGT92 working again, and it’s sitting in my office and getting -65 db or so, where before it definitely didn’t get that high before. So I’m very interested into why i’m getting better signal with this $100 TTN white box than a $150 PI Hat with a huge heatsink…
Would definetly suspect connections then - the chipsets are broadly same and both known good bits of kit from design, build and construction perspective. another possibility is if you have induced some static damage on the RAK front end - enough to degrade but not to kill completely - have seen conc cards degrade 10-25dbm but still work in such cases…but 1st off inspect and double check that ufl connection etc. re-seating as you switch units may be making a difference… gently manoever cable angle (basically rotate around the connector) and monitor regular packets to see if there are any chages in RSSI - also what SNR values are you seeing in the areas you are testing (not just RSSI)…
I wanted to provide an update. Turns out it was due exclusively to me misunderstanding the fundamentals of antennas. Once I learned more about how the high gain antennas worked, I realized a huge issue with my location. I may not be in a super hilly area, but it’s enough. Basically to my south (which is where I was doing the majority of my testing…) there is a hill that tops off higher than the combined elevation of my house + antenna. So the beam was driving straight into the ground. I occasionally got a packet from up and over that hill, but it was only when I was on a slightly higher elevation. But once I went back down, gone. I was able to get 1.1 km using my LGT92 and it was pretty low rssi (around -120) But it was consistently sending data. This location is downhill from my house, and even though there are trees in the way, it still was able to get the signal.
So the lesson learned is when I was testing from the ground, I was too low to get a solid enough piece of the beam. I found a street a block away that was the elevation of my house + antenna. I got -63 rssi which was the best signal I ever got with the LGT92 outside of standing right next to the antenna on a ladder. There were definitely trees in the way and a house or two, but I still got great signal. I drove around in a neighborhood to my east that was not blocked by a hill, and when I was on top of hills, I got signal, but then as soon as I dipped even a little, it went away.
So now I’m realizing there is probably nothing wrong with my gateway, just how I installed my antenna. I was so heck-bent on using the existing satellite TV mount that I didn’t think about how low it was on my roof. So i’m re-thinking my antenna mount, and going to mount it right on the top of my gable. It should match the hill to my south and hopefully allow signals to hop over it. And it should cover a couple more km for sure.
And also worth remembering that fitting a ‘high’ gain antenna does require you (in most places in the World) to step back the transmitted power to keep the gateway legal.
So the net effect is no actual improvement in range but a restricted area of coverage.
Paging my newly discovered expert on exactly this topic, @snejra, this is why we need some training videos from you!
And here’s a question for the likes of @Jeff-UK who has probably seen more antenna setup than the rest of us!
I have just built a RAK2245 + BasicStation + RPi4 + PoE hat to replace what has been a really superb V2 gateway. The old setup used a RAK831 concentrator and regularly gets packets from Wales to where I live in the East of Ireland. The antenna connectors are different however!
I know that the antenna as it is works really well even with 5m between the gateway and the antenna on the top of my TV antenna pole.
Am I better using a UFL to SMA tail and connecting to the existing SMA or should I replace the connector on the RAK2245 which sould be a really simple soldering job, all the pads are there?
What sort of injection losses and cable losses will I get with a UFL to SMA? Will I change the tuning of the antenna path by changing the connect or to suit the existing cable straight to the concentrator board?
Anybody care to volunteer some wisdom?
Thx Garry
Hi Again Garry, Personally I would not bother as board is laid out for the HiRose u.fl - not sure on the ground lands and placement on edge of the board (to side of the ufl) whether a straight on SMA would work effectively, however, the risk of damage and mismatch (plus void warranty!) not worth the potential reward - you might gain 0.1-0.5db through change but a bad implementation could cost as much or poss loose you the board! Note I dont know what your expectations are but both that and the 831 are SX1301 designs and very similar from RF /Baseband performance perspective other than the physicals and heatsinking etc. so you wont gain much with the change I dont expect. (IIRC the 2245 is a dual SAW implementation which will carry its own insertion loss so little/no improvement in RSSI sensitivity (again IIRC), though you might see some improvement in proximity bands from adjacent noise reduction and hence a marginal SNR improvement? Indecernable in most installs I suspect!). My pref would be a very short pigtail to get from ufl to SMA asap, pref even go ufl to N-Type in short order. Then if you want to gain a little I would improve the ant feed - dont know what you are using right now but at 5m worth spending a little to improve - e.g. look at some of the Aircell ultra-low loss cables or equivalents - I’ve had custom cables made up in the past for SMA (smaller gauge Aircell - good for 3-8/10m) or N-Types (thicker gauge => lower loss) with good results for 5,7 10, 12 and even 15m runs. Note: if you do risk change be sure to solder the SMA on both sides of PCB as helps with strain relief (another advantage of using a flexi pigtail of the ufl) esp if connecting directly to a ‘strong’ cable type. A lot of manufacturers often solder 2 of the 4 SMA corner posts only on one side and I’m afraid I include RAK in this (have had RAK5205 SMA field failures in high stress use cases as a result!). C/w cable losses the local connector losses and reflections will likley be small … and what the hell you already hear Wales! (Remember the wider/further you can hear the higher your contributed noise floor and potentially reduced/impaired SNR capabilities)
I tend to agree with you @Jeff-UK but only on my gut feeling, you have been around the block a good few times! Many thanks for the detailed feedback too 
I do have a u.fl to N-Type tail which is approx 1m long and a N-Type RAK antenna (maybe 500mm figerglass tube which I believe has a tuned pcb within). I was not over excited about this particular antenna when I got it and other reviews tend to concur! I have never measured standing wave so won’t say anything bad until I do but there may be a case to extend my existing PoE that reaches the bottom of a 5m pole to a gateway mounted just below the antenna with that u.fl to N-Type connection.
I should do some crude and not very scientific testing to see what sort of RSSI results I get with the all RAK hardware, that being the PoE powered RPi4 with RAK2245 hat, RAK supplied u.fl to N-Type tail and then the RAK antenna and see how it performs.
Anybody any suggested packet forwarder favorites for a rebuild of my existing two RAK831 gateways? My main tester is on a RPi4/PoE above my house and I also have a “Charles Hallard” Pi-Zero-W fallback also using a RAK831. I’m kind of liking the Balena Docker I use to manage my main tester still on V2 and my new V3 we’re talking about above.
Suggestions welcome? Anything on Balena? Does the RAK831 work with Basicstation (pin mapping accepted)?
Thanks again for the support
G
Not really sure why you’re changing gateway hardware at all vs putting in V3 settings and maybe updating the packet forwarder or switching to a new scheme like basicstation.
Anyway, the only real difference between the RAK831, RAK833, and the more recent RAK2245, RAK2247 etc is form factor and that the latter have a slow, unnecessary SPI level shifter. That requires reducing the SPI clock speed, officially I think 2 MHz is supposed to work, but I run everything at 1 MHz, even though most of the time it’s still actually the older cards installed.
It’s only the sx1302 (and I guess sx1303) as in the RAK2287 where you get to a need for fundamentally different software drivers.
Way too long and a waste of signal as cable very lossy…I was thinking more 3-10inches max ~8-25cm?
A few updates on this Pi hat and my old V2 RAK831…
I have the RAK2245 stacked on a Pi4 with PoE hat sandwiched in between and it works a treat so far. I used the best quality short u.fl tail to join to the existing antenna. I live between low hills and the sea on the east coast of Ireland. The local town is in a hollow so no coverage. I know folks hit my gateway from Wales as they tell me they do. Have a look at this metadata from the V3 Gateway console. I am thinking this came over the Irish sea for sure. SF12, RSSI -123, SNR 19.5 says this made it by a whisker!
"settings": {
"data_rate": {
"lora": {
"bandwidth": 125000,
"spreading_factor": 12
}
},
"coding_rate": "4/5",
"frequency": "867300000",
"timestamp": 1831410132,
"time": "2021-10-13T23:46:20.206509113Z"
},
"rx_metadata": [
{
"gateway_ids": {
"gateway_id": "ttn-v3-rak2245",
"eui": "DCA632FFFEC89168"
},
"time": "2021-10-13T23:46:20.206509113Z",
"timestamp": 1831410132,
"rssi": -123,
"channel_rssi": -123,
"snr": -19.5,
"uplink_token": "ChwKGgoOdHRuLXYzLXJhazIyNDUSCNymMv/+yJFoENSzpOkGGgwIzN+diwYQy87NjAIgoMizxKbaRw=="
}
My old RAK831 is also on a Pi4 with PoE and now runs basicstation firmware on a balena docker image. This just needs an enclosure and an antenna to put back into service somewhere. I am happy to sponsor a school deployment if any STEM teachers in Dublin/WIcklow ever look in here?
G