In the "Messin' With Texas" paper, there was a suboptimality in the way I calculated the risk of MMN compromise. So, I wrote up a page pointing out the suboptimality and what it should be instead. This correction does change the graphs in the paper, but the correction does not change the major results because at the most vulnerable end (only allowing 1 or 2 guesses), the suboptimal measure and the optimal measure are one-to-one.

Link: Security: Errata for "Messin' with Texas - Deriving Mother's Maiden Names from Public Records"