But I haven’t used a network vector analyzer to optimize the output. The pi matching, if necessary, will be specific to the particular pcb design, materials, and components you are using anyway. You can’t just copy a matching circuit from a random board and expect optimized results. It really needs to be tested to fully optimize. But for the kind of relatively short range (< 1 km) use I make of LoRaWAN, even an 82 mm copper wire works quite well as a 915 MHz LoRaWAN antenna.
The chip/pcb antenna on Grasshopper has been tested to work well for at least at 1.4 km distances. I suspect is would continue to work well in the open out to 10 km or more.