I believe comparing FSD to human driving is kind of pointless. However, FSD safety performance is vital for regulators. As long as Tesla can show them people are using it responsibility which the data obviously display, then it keeps regulators off their backs and there wouldn't be some kind of Dan O Clown's wet dreams coming true with a nation wide ban.I find these stats a little disingenuous. FSD by the very nature that it is still iffy means, not many are turning in on, and those who are using it are doing it with extra caution and taking control when it makes a mistake (which is good). So these stats are not a reflection of how good FSD is, rather how much attention it needs.
The same reason why there are few accidents on the L2 driving tech from other makers compared to Tesla. Because they all suck, very few use them and those who use do them only in very safe places, whereas Tesla drivers use Autopilot anywhere and everywhere and are completely lulled because it is so damn good.