We should graduate the best engineers and scientists.
But if you have a bigger pool to draw from, the 10% that's not crap will amount to more engineers. And the 0.01% that's really spectacular will be more spectacular. That's the way a bell-curve works - if you raise the area under the curve, the tips spread out as well.
There're a lot of smart people in finance whose talents are being wasted making sure that the market reflects all information within microseconds instead of within minutes. If the economic incentives were such that they started companies or did basic research instead, I bet you'd see quite a few more breakthroughs.
I don't know what the right answer is (or if there is one); what feels right is to raise the standards of admission.
Your second point assumes that the relative percentage of "not crap" remains constant as enrollment increases, which is highly dubious.
And there's a lot of smart people being wasted at Google and Microsoft, too.