In hindsight, Monterey Dunes was a poor choice for a study that compares
Ookla Speedtest.net and CalSpeed. Coverage in this area is basicly the
cellular equivalent of fumes. Radio diffraction and self-interference 
provides what appears to be random peaks and valleys in the coverage.  
Handsets work hard to find a signal, repetitively resetting their radios 
and switching between 3G and 4G.  This appears to cause colateral damage 
to GPS information and makes measurements appear "all over the map." 

Our field survey instructions call for annotating a paper map with 
measurement time stamps from the tablet's clock which is always displayed.
Here we had to wholesale replace the GPS coordinates for both CalSpeed
and Ookla with coordinates picked from the map using the time stamp as
the common key.  Both test programs record the radio in use (LTE, Ehrpd
or Wi-Fi) at a point in the test.  But if the handset switched radios 
in the middle of a test, this is not recorded and the indication can
be misleading.

That said, that the agreement between CalSpeed and Ookla measurements
is good is not a surprise.  Under the hood, they are both using iperf
with multiple streams for their TCP measurements. CalSpeed reports
successful and failed tests; Ookla records only tests that complete 
successfully.  Ookla failures can only be picked up from notes that 
the surveyer leaves on the paper maps, and this is an imperfect 
process -- especially when the surveyer is running two surveys for 
comparison but carrying a single map.