When I checked the (full sized) CAN PMU circuit diagram it looks good and the voltage is measured from the battery input side of the current sensor. The CAN PMU Lite might be different, where the voltage is measured from the drain side of the current sense resistor - this would explain what you are seeing.
I found exactly the same issue with the CUAV HV PM and I made a simple voltage divider and connected it to the battery input side and output to the voltage sense wire (obviously breaking into the voltage sense wire so it’s not getting the CUAV voltage reading too).
This could be much harder to “fix” with the CAN PMU, as you’d need to do a bit of circuit board surgery.
It might be easier to just set up a voltage divider and feed it into the analog voltage sense port and set up BATT2 with a suitable voltage divider value and battery failsafe parameters.
Take the battery voltage failsafe parameters out of the 1st BATT group and just leave the current measurement in place.
This is disappointing and CUAV should really fix those PMUs to read the voltage accurately under load. To fix it ourselves makes a mess of the wiring and we fiddle with the parameters, when it should be neat and “plug and play”
EDIT
For 6s Li Ion use:
BATT_ARM_VOLT,19.10
BATT_CRT_VOLT,18.00
BATT_LOW_VOLT,18.60
MOT_BAT_VOLT_MAX,24.60
MOT_BAT_VOLT_MIN,16.80