I read somewhere, that some power modules uses a shunt resistor to measure current draw.
Thus the GND on the motor side of the PM is seperated from real GND by the resistor, and we will have a floating ground at the ESC. Connecting the ground wire from the ESC -> APM/PX4 would act as a an almost short (everything has a resistance), with the shunt, giving wrong/no amp draw values.
The recommendation I have seen is the remove the GND wire from the UBEC of the ESC connected to the throttle channel (the one that powers the servos).
But then we can expect the UBEC to drop voltage as motor load increases, as the motor resistance will decrease as amps go up (not really, as it is constant but pulsed, but averaged over a period it looks like it). The closer the motor resistance gets to the shunt resistance, the more voltage will be dropped over the shunt. It might be only 0.1 volts, but it depends on the shunt vs motor load/observed resistance.
If we short the shunt, we will get stable UBEC voltage, but no amp draw measurement.
I think this should be in the documentation.