It is very likely the Basques, etc., were once predominantly G2a and have only become mostly R1b via admixture, as they have come to resemble their predominantly R1b neighbors. We see a similar phenomenon among the Ossetians, who speak an Iranian language and were probably once mostly R1a, but who are now mostly G2a like their Caucasian neighbors.
You can't really use the frequency of a y haplogroup in a given area to judge whether or not it came from there or was once very frequent there. One has to look at haplotype variance and the SNP trail. Judging strictly by frequency, one would expect all kinds of R1b to be showing up at Neolithic sites in western Europe, yet that is clearly not happening. The oldest R1b yet found in Europe dates to about 1,000 BC (Bronze Age).
The tiny percentages of R1a found in western Europe are mostly explicable by historic references. Why strain credulity by explaining them as "kurgan elites"?
The Romans, with their advanced civilization and advanced administrative, political, and military systems were unable to impose their language on all of their empire. The Germanic barbarians, the Vikings, and the Normans were even less able to impose their languages on the people they conquered. How likely then is an "elite dominance" model for the transmission of early Indo-European languages? Not very.
Ever notice how horse-riding steppe nomad incursions throughout history fizzle - linguistically, genetically and otherwise - at the Hungarian Plain?