This is why no one can trust libertarians to analyse risks rationally.
You're "not sure how that would happen" but there are decades of studies showing exactly how it does happen, who the victims are, and what the quantified risks.
The primary risk is to other people inside the car, then side ejections. Front ejections are a footnote.
You decided only the last of those is a problem without considering the other possibilities.
When considered as a whole, the evidence is absolutely clear that set belts save lives.
It's the same story with vaccinations and other mandates. "I don't like being told what to do" turns into "Well, obviously, the real problem is..."
The people die unnecessarily in large numbers - far larger than if the measure really did cause mass harm.
But how frequent are those events? I'm happy to be wrong, I just never saw it as a likely or common occurrence and for me it falls below the level of risk with which I want to empower the government to regulate it.