Are you able to estimate from your results what Holman et al.'s real misclassification rate is? I'll wager it's on the order of 0.5-1%, but it'd be great to know for sure. I know 372 unique names is not many, and the original plan was to contact authors to collect more manually assigned name-gender datasets, as you have done here. However, the groups I contacted for data did not respond, though they have provided data for this study. ¯_(ツ)_/¯