Aside from the “you can’t compare those numbers across categories” issue, there’s the fact that Tesla makes the incorrect inference “safest per collision” ⇒ “lowest probability of injury of all cars”.

I’m not making any claim about Tesla’s cars (because of its low center of gravity and relatively large weight, they likely are quite a bit safer than the average car in collisions) but if brand B has distracting UI, bad steering, or whatever, that causes it to slam into walls twice as often as cars of brand C, being safer in a single collision may not offset that.