"...I learned a whole bunch of fascinating things: about the Doutai speakers (NW New Guinea) who suppressed their language's implosive consonants for several weeks while Mark was studying their language (a phenomenon reminiscent of Dan Everett's experience of living three years among the Pirahas in the Amazon before they stopped suppressing their linguo-labial stops in talking to him); about plugging in typological features — word order, consonant types, etc., etc., etc. — to biologists' statistical models and coming up with areal rather than genetic groupings in known cases (e.g., Rumanian grouped with Slavic rather than with the rest of the Romance languages)..."
To what group a language ends up belonging depends a lot from what "typological features" are plugged in. The Warnow, Ringe, Evans, Nakhleh approach, using a Swadesh list and a few morpho-phonological features, put Albanian in quite a variety of linguistic neighborhoods. However, the list of features for Albanian was not very representative and seemed to be constructed from the Toskë dialect instead of Gheg which has more archaic features). I remember that there were a few dubious words in the Albanian Swadesh list (e.g. claw is thua instead of kthetër, pierce is ther instead of shpon, etc.).
I hope the Romanian list of features was better constructed.