I am sure Paul will follow up (after all cellular networks is his day job!)
My understanding is that the footprint for the cell transmitters is not designed for airborne use, and is therefore sub optimal.
In the early days of testing the OGN-R data in the air, we also connected an internet feed (via 3G/4G) to the spot the glider site - the latency was completely unacceptable. Now this could be due to filtering in the software, but I got the impression it was purely down to the data refresh rates.
Thx
Lee