This data needs to be taken with a grain of salt. For the longest time FSD was geofenced to highways and other roads that are easier to navigate. And nowadays people are likely using FSD when it is reasonable / safe. This is why FSD likely has a statistical advantage because of when/where it is used.
Likely the only comparable data will come from those robo-taxies.
But the basic premise is sound - strict adherence to road rules by an autonomous system should decrease accidents.

Ahab!

I am sure the insurance company is trying to lose money and that is why they are discounting the "third party data". /s

Too bad actual third party data shows the literal opposite, lmao.