I think that the coolest part of those findings is that we can actually start reconstructing Proto-Italo-Celtic and Proto-Graeco-Armenian with some certainty, without finding too many spurious patterns. That would be specially useful to reconstruct the Proto-Indo-European stops; I'll show it with a table, with the stops of the current PIE reconstruction and the typical reflexes into four languages representing those branches:

The first thing that I notice in this table is that Armenian is satem and Greek is centum. And yet they're more related to each other than to their respective groups? This means that the centum/satem division was never different IE branches, it was simply an areal feature.

For most part Latin transformed the so-called "breathy voiced" series (third line) into a bunch of fricatives; in some situations they became plain voiced consonants, but I didn't add them to the table. In the meantime the Celtic languages consistently merged that series into the plain voiced ones. The difference in outcomes shows that Proto-Italic-Celtic still somehow kept the distinction; and this makes me suspicious if that series is composed of breathy voiced consonants at all, given the lack of symmetry I'd expect them to be relatively short-lived. Greek vs. Armenian makes it even more puzzling - for Greek that series is the one with the biggest voice onset time, but for Armenian it's the one with the lowest!