It wasn't that by the end they didn't understand the problem, but rather that by just attacking "build a human powered plane that can do <x>" they were not making progress. Realizing that in order to "make a human powered plane that can do <x>" they needed a lot more data and their development method wasn't getting them that, they changed to "make a human power plane in such a way that I can get a lot more data quickly". I'd assert that realizing that the problem is getting data too slow as opposed to "working" too slow is not one that's immediately apparent to most humans.
Got the same impression. The goal was X, realized that you first needed to do A,B,C before reaching X. Isn't that something we all do intuitively or intentionally all the time?
I agree with tintin - it's not that the wrong problem was being attacked, it's just that the methods of solving the problem weren't very good. It's another example of protopying, and iterative and agile development beating BDUF.
The conclusion is: to solve a problem you need a tool with which you can iterate fast.
The author states that we don't understand we need such tool...