MB Drive Pilot is designed not to attempt to operate outside of a well-defined domain.
It will, by design, not operate in a way that is unsafe or in a context in which it's unsafe.
You are arguing the system would perform worse in contexts in which you have not seen it perform (because its designers decided to prevent it from such attempts).
What evidence do you have for your argument?
Restated: You are assuming that if Mercedes engineers had a similar appetite for putting millions of innocent people at risk by operating in unreliable contexts, that their system would perform far worse than FSD does. On what basis do you make that assumption?