If the governor is operating like CastleCreation ESCs, then the rpm curve is flat which means that the motor curve fit parameter should be adjusted so as to let the flight controller know that the motor response is "linear".
Assuming that in the case of variable pitch, motor curve fit applies to throttle and not pitch pitch, my rough guess is that if you find the proper motor curve fit, the PIDs calculated by the autotune a low throttle would be good for high throttle. Don't take my word on that and let the developer confirm my assumption
Have you tried to autotune the aircraft ? Are the resulting tunes vastly different? Maybe you should tune the aircraft and share the tunes at low (3200) and middle (4000) rpm. Start first with low rpm, if the aircraft shakes a lot a low rpm, reduce "autotune actuation effort" in system tab, if it barely move you can increase the effort a bit: the objective is to have an the aircraft gently wiggling during the autotune session. When the tune works at low rpm, try to tune at high rpm. For each autotune do not forget to "share the tune with the community" (there is a button for this). Once we've got both set of tunes we can see if we really need to dynamically tweak the Roll-P and Pitch-P with the TxPID module.