... The charge a designer gets when running a site through the W3C's validation tool. Great, now this site is squeaky clean and there's not a thing wrong with it! This is probably never true.
Jason Kottke has written up some very interesting thoughts on web standards and semantically rich documents. He has many valid points (sorry). But seriously, it's a problem.
When designers see using valid XHTML as a religious activity, there's bound to be little or no semantic structure to the document. That said, I can agree with Doug that creating pure semantic structure while at the same time making it look great is a challange that's nearly impossible. Given the current state of CSS support anyhow.
But that's where the balance comes in, and I'm reminded of Tantek's A Touch of Class post from way back when. I've found that one of the biggest problems with newly authored XHTML is the overuse of the class attribute. The document might validate just fine, but there's a class assigned to just about every element in there. Not to mention the possibility of table and font tags for presentation that Jason rightly points out will still validate despite the fact that they may be littering the code all over the place. I also attempted to comment on the subject back in April.
The validator is dumb (literally), and unfortunately is sometimes used to validate the designer first and the code second. We don't have a semantic validator and probably never will, but the smarter we get about structure, the closer we'll get to seeing improvements.
Like Dave, the best thing I could recommend to any designer/developer looking to really understand that balance is to pick up a copy of Jeffrey Zeldman's new book Designing With Web Standards. Read it. Live it. The guy knows what he's talking about.