The study you cite shows an opposite result.

Are there independent replications that actually show the same effect? I recall an attempted replication that produced equal miscalibrations on the upper and lower ends of the distribution, suggesting that the effect was just mean regression.