That's what is weird about this: from the paper, "Unlike population age, we find no significant correlation between host metallicity and Hubble residual". So it correlates with age, but not metallicity!