. . . Says 18% of the Free State could speak Irish (no definition on how well) in 1926.
So people who identified as Irish speakers in 1926 didnt' have a cúpla focal dimly remembered from a primary school Irish class; most of them were people who spoke, or had spoken, the language in a domestic environment. And that makes for a much higher degree of competence.
It's true, there were people who had taken Gaelic League classes for ideological/political reasons (or because someone they fancied was taking the classes, or whatever). And even after a couple of years of Gaelic League classes you might have very limited command of the language. But Gaelic League classes had a limited reach; certainly nothing like 18% of the population.
So my guess is that the great bulk of the 18% who recorded themselves as speaking Irish in 1926 probably could speak reasonably good Irish, and did speak it or at least had actively spoken it at some time in their lives. Whereas now we have a very large class of people who studied Irish academically at school but have never used it at home or in the community, and who have pretty limited Irish.
I suspect that the truth is that the proportion of the population that knows some Irish is much larger than it was in 1926, but the proportion of the population that has functional Irish - that could engage in a simple conversation, say, or read a newspaper article in Irish - is probably smaller than in 1926.