If you do a little math, you can convert that into what fraction of the total books read by the group as a whole are read by each slice.
Depending on how far you want to take that 21+ slice up, the 21+ group is reading about half the books (a couple dozen per person), 2/3rds of the books (50 a year) or more.
So this presents a conundrum. Why are commentators expecting a plateau in e-book content at 35% of the titles sold? Either power users aren't fully converted (this is trivially true, for some small percentage -- I know one of my regular readers and my walking partner have not yet converted to e-readers), or the plateau shouldn't show up until some higher point.
There's another possibility. Some fraction of book content is _still_ only available in p-form (I know, I've had to acquire several books in paper in the last year and I'm not happy about it; usually these are academic monographs). If a large fraction of power readers are reading a lot of books that are not yet available in high-enough quality e-form, then they won't convert. Period. At least until the quality problem is solved. But the survey data is specifically about trade books, so that doesn't seem too likely. It's also possible that you have conversion averse buyers (parents, grand-parents, etc. who don't read much themselves) buying p-form for power readers who would prefer e-form but aren't going to make a big stink about it.
Or the data is bad. Yeah, that'd never happen.
Here's how the data could be bad:
If indie ebooks are not in the projections or the universe under analysis, and they are a quarter, then total ebook sales would wind up plateauing somewhat higher than 35% or 40%, and might get a lot closer to the 50% or so that would be the low end of what seems reasonable to me.
ETA: If you are blinking, and thinking, walkitout, libraries! Well, you are clearly smarter than me. Libraries could, indeed, explain a lot.