> So the skills that they should emphasize in graduate schools
> are those of a used car salesman not that of a researcher.
I think this is extreme: used car salesmen can't design cars or even build them. It certainly is the case that knowing how to "sell" your reseach is important, but you also have to be able to do it. My grandmother used to say that if you can't explain something, you don't understand it. The difference between the two levels in academia you present (demi-gods and the rest) is not typically "only" salesmanship.
And frankly, I don't think any of this is opaque to graduate students: they see this issue 24x7. So if they don't pick up on it before they get their degree, they aren't paying attention. I doubt seriously that you somehow were taught the wrong skills ... even the process of picking an advisor and working to "sell" your thesis to your committee should give you a solid sense of the structure of academic research. And then, of course, you have to do the research and write the paper.
> I believe this is true of most workplaces in the US - producing
> things will not get you very far, selling them will.
The biggest problem I have seen in academia (I've worked closely over the last 25 years with research groups from Columbia, Stanford, UC Berkeley and CMU; less closely with a few others -- so this is likely the "best case scenario" ... I can only imagine it's much worse in other places) is that while both of these skills are important, very few researchers feel it's useful to collaborate to any great extent. Even the best that I've worked with (who are, by extension, some of the best in their fields) are loathe to work together with others who have complimentary skills and experience. They feel as though they have to go it alone.
Which I think explains why so many of them are lousy at it.
/jordan