This is confusing: Tesla publishes that Tesla's vehicles, both with and without autopilot, are the subject of substantially fewer accidents per million miles dr

This is confusing: Tesla publishes that Tesla's vehicles, both with and without autopilot, are the subject of substantially fewer accidents per million miles driven than the US average; last quarter, somewhere ~4.8M miles/accident on autopilot, ~1.5M miles/accident Tesla w/o autopilot, and ~0.5M miles/accident US average. Given I'd like to assume the best and conclude no one is lying here; obviously what they're measuring is slightly different, but it feels like they're similar enough that we shouldn't be getting entirely opposite outcomes. Do Tesla drivers drive substantially more? Are Tesla drivers just more likely to report crashes to their insurance/police (versus, Tesla's data is probably from automated internal observability)? If Tesla's data is based on internal observability, is that observability underreporting incidents to the mothership? [1] https://www.tesla.com/VehicleSafetyReport https://www.tesla.com/VehicleSafetyReport