I wonder, sometimes, how true that is.
I went into the startup world for exactly that reason: I wanted to start my own company eventually, knew that I didn't have the knowledge or experience to do it immediately, so I figured I'd learn the ropes on somebody else's dime. I picked companies that would give me a lot of responsibility, and ones where I could directly observe what the founders did and why they did it. And I took careful mental notes whenever something worked well or didn't work well, and resolved to do the same thing (or not do it) when I started my own company.
And then I actually did start my own company, and found that I made nearly every mistake that my old bosses had made, I just made them in new ways, and stopped making them a little quicker.
I found there're three main problems with the "learn how to start your own company by working at someone else's company" school of thought:
1.) There are many more ways to fail than there are to succeed. Therefore, if you pick any random strategy out of a random company, and you don't yet know whether the company is failing or succeeding, it's probably a failing strategy. Orson Scott Card once said, "There are a million wrong ways to tell a story, and a thousand right ways." That's probably true of entrepreneurship too, just multiply the numbers by a few orders of magnitude.
2.) If you haven't yet worked at a successful company, you probably don't know how to tell the difference between a successful strategy and a failing one. A lot of the things I respected most about my past employers turned out to be mistakes. For example, one of them was big on code ownership, and seeing how hard coders worked at getting their code running, I thought it was a great way to encourage people to take pride in the code they produce. Another was building a platform for financial apps, and I thought they were geniuses for their strategic insight - after all, they get recurring revenue and their customers can't easily switch away. It was only after I tried building a platform myself that this reached the top of the list for "things startups should never do" - and given the number of YC startups that get funded with their own webapp platforms, I guess lots of other founders haven't yet learned that either.
3.) Many times mistakes don't look like mistakes, because the circumstances that surround them are different. Very occasionally, they aren't mistakes, because the circumstances that surround them are different. For example, while I was building my platform to create casual games, I thought I'd sidestep all the complexity & adoption problems my financial employer had faced, because I was purposely limiting the scope of my platform to games and wasn't trying to have it do everything. I found that no, the platform part really was the problem, and no matter how much I polished it up, that approach just didn't work.