This kind of reasoning is worrying. On one hand, based on the high failure rates, it can be argued that start-up founders are insane. On the other hand, rationality shouldn't automatically preclude you from taking high-risk, high-payoff bets. Another rational way to approach this would be to actually figure out what separates successful start-ups from the failures and then just do that.
The few sane people that make startups usually succeed, the less sane people make startups and usually fail, that's my theory.
Basically, there are the sane people who have a good reason to think they succeed, and they make startups, and they succeed (usually), and there's also much larger number of people who don't have any good reason they'd succeed, and out of those people, those who are less sane, try to make startups, and fail.
You can't generalize the sanity of two vastly different populations (those who fail and those who succeed) from average.