You can titrate the amount of startupness you get in your job by
the size of the company you join. If you're the first person hired
by a startup with one or two founders, you'll probably be a de facto
cofounder. Whereas if you get hired by a startup with 30 people
you'll have a lot less risk and stress, but it will probably still
be more interesting than going to work for a big company.
Another reason is that luck (usually in the form of timing) is a
big component of startup outcomes, so in some cases it can be a
better bet, measured both financially and by how much effect you
can have on the world, to join a startup that is obviously taking
off than to start your own. All the first several hundred employees
at Google, to take an extreme example, probably made more money and
had more effect on the world than they would have on average by
starting their own companies.
Plus you need more than ability to start a
startup. You also need an idea and probably a cofounder. Joining an existing
startup can be the easiest way to get both.
Business is a very particular discipline in its own right. While the technocrats and bureaucrats of corporate America's professional management and executive corps have certainly done enough to exaggerate its exclusivity and idiosyncratic character, the fact remains that it's a rather distinct kind of thing. The qualities that make one a good programmer, a good technologist, a good innovator, etc. aren't necessarily the same qualities that make one a competent or happy business owner, especially when, as a matter of practical necessity, much of the everyday workflow is tedious administrativia or other things banal to engineers.
I'm all for the idea that intelligent, confident and well-rounded people are better off doing something more interesting than fighting mediocre fights in the cubicle trenches, but that doesn't mean they're all--or even a majority of them--are entrepreneurially disposed per se.
Plus, as has been pointed out elsewhere in this post, joining early often brings equity with it, which, from a certain point of view, is quite an entrepreneurial proposition.
If nothing else, it's simply not mathematically possible for everyone to be the chief in the driver's seat; many technical endeavours seriously require people to build out and scale.
I think this is my favorite metaphor ever. (I'm titrating it all the way down to homeopathic levels myself...)