It’s great (a) to find a study group releasing its data set so we can all take a look to see what the data is really telling us, and (b) that you’ve gone to the trouble to walk through the set and explain some of the finer points.

From personal experience, as someone who’s got very little facility with foreign languages, I can say anecdotally that the “time spent = probability of fluency” equation is correct. I taught myself French starting in my 30s and reached an adequate level but the last 8 months living in le monde francophone has enabled me, thanks to speaking the language every day in a wide variety of contexts with a wide range of different people, to become more-or-less fluent. Likewise my modest Russian language skills benefited most from the times I’ve lived and worked in Russia and Ukraine (where Russian is still the dominant language).