The submitted URL was correct but, as detaro points out, the canonical URL wasn't, and our software now uses canonical URLs when it finds one.