A way to measure filter offet and get a good value is to make a sequence with just 2 filters, L and a different filter, set event order to rotate through events and have autofocus to run during filter changes.
Do maybe 10-20 exposures for each filter at a very short exposure as exposure time doesn’t matter.
After you’re done running this you can open image history, note down all focus positions, remove any outliers, sum them up and divide by the amount of datapoints you are left with for each filter.
You are left with 2 numbers where you can just subtract one number from the other and be left with a pretty good filter offset value.
Mabe it would be possible to add a filter offset trainer in SGP that does this automatically?