I dunno. I learned relational theory in college along with SQL. Since then, I've forgotten all of relational theory since I never used it for anything, but my SQL has improved quite a lot through use.

I'm thinking you may have a third variable problem here; people who write sucky SQL tend not to have spent a lot of time thinking about how to write non-sucky SQL. People who write non-sucky SQL have. The former has nearly zero chance to have been exposed to relational theory...the latter has a non-zero chance to have been. Correlation /= causation and all that, and the actual causative variable is time spent learning how to write non-sucky SQL.