D-K claim a non-monotonic relationship, which simply isn't supported by that data, as you yourself point out: people rank themselves correctly (ordinally). I didn't mean to say that all self-assessments are the same, if that was the misunderstanding. My point is that the self-assessments indeed are meaningful, even more so than D-K claim.
Check the original paper by D-K. Fix only focused on the first plot which has a monotonically increasing trend. The later plots show varying degrees of nonmonotonicity, though sadly they don't include error bars to indicate how statistically significant the differences between groups is.
