My uncle is an aerobatics pilot, and he says that the main reason aviation is so safe is due to the meticulous and ubiquitous use of check-lists. Every time there's an accident the intent is to analyse it so we can update the check-lists to prevent it from happening again.
You're right, we don't know what a successful startup should look like, but they didn't used to know what a safe 'plane was like. Because of the unrelenting analysis of crashes, we do now.
Not only is the problem space nearly infinite in comparison to flight, but the "physics" governing business success change impossibly fast.
None of that is to say that we shouldn't study failures, but the question is which failures have widely applicable lessons? If you indulge me to abuse the analogy further, most startups are like those early flight attempts by people who jumped from high places wearing ridiculous balsa and fabric contraptions.
I think the analogy might best be put at the borderline between flying-chicken-man and the Sopwith. Obviously there are some folks that are flying, and flying reliably, obviously some things work and some things don't -- so now is the time to start writing it all down and sorting it out.
The problem is that we can't even see the airplanes of the folks who are flying -- we just get to hear the cool stories about how they got from point A to point B.
I look at places like DreamWorks which are able to capture "lightning in a bottle" over and over again or some of the serial entrepreneurs out there, and think that it is possible to take highly variable environments and put a little bit of structure around them. Nowhere near as much as aviation, but like the other posters pointed out, the only reason we have that for aviation is that a lot of smoldering craters became somebody else's life lesson.
Well, the reason this works so well for airplanes is that the airplane used to work, and everybody knows exactly what a working airplane is supposed to look like.
In general, nobody has any idea what a working version of a failed startup would have looked like. There are too many ways to succeed, and so many chaotic events that affect the timeline in profound ways (imagine if the inventor of CP/M hadn't been out when IBM came to call...) that you can't really think about the various alternate histories without quickly running into a combinatorial explosion. So I think your metaphor is not very useful. ;)