Would you say that more recent methods allow us to talk about degree of language similarity in a way that is objective enough and meaningful/useful enough that the results are not "for entertainment value only"?
"Objective enough" isn't the issue, rather it's how confident we are in the inferences. And yes, I would say that the new methods are preferable because 1) they do a better job of explicitly quantifying and representing uncertainty 2) they use more sophisticated source data/features.