While the Tesla safety report [0] does compare Teslas with and without Autopilot, and the safety rate does look to be 3-5x better in terms of millions of miles

While the Tesla safety report [0] does compare Teslas with and without Autopilot, and the safety rate does look to be 3-5x better in terms of millions of miles per accident, it's not clear if the stats account for the fact that Autopilot can only be activated in ideal conditions, e.g. highway driving in relatively good weather. In other words, what is the accident rate of non-AP Teslas on highway/freeway only versus the Tesla-AP rate? I'm assuming Tesla doesn't have an easy way to figure out weather at time of accident (of course, wouldn't be hard for a dedicated researcher). But accidents-per-highway-miles only shouldn't be burdensome to calculate and would provide a much more apples-to-apples comparison [0] https://www.tesla.com/VehicleSafetyReport https://www.tesla.com/VehicleSafetyReport