But then you reach a contradiction: 99% of the population cannot possibly by literate in the strict sense ("being able to decode a sentence at age 15") if the other statistic of 1 adult in 6 being unable to decode anything above a fourth grade level ("decode a sentence written for a nine year old") is true. And it's not simply a matter of "ignoring results you don't like", since 99% literacy is not borne out in our day-to-day experience -- Americans don't read nearly as much as Scandinavians or Japanese do -- yet, if the US only had a 86% literacy rate, the US would not be able to maintain a first-world economy. A statistic that relates intimately to social morbidity shouldn't be subject to this much discretion.
One measure is the % of Americans who are older than 15 and can read at any level.
The other measure is the % of Americans who are older than 18 and can read English at a level of moderate sophistication.
It should not be surprising that the second measure is substantially lower than the first.
In which case, the first measure is the less accurate one, since a fifteen year old who reads poorly will not be any more proficient at fifty without extensive remediation.
I think one reason for the discrepancy is that literacy screening at age fifteen may not include dropouts. The
high school dropout rate is well studied (and has been declining for twenty years), but are
middle-school dropouts accounted for as well? If the stats for literacy are from the 1980s, then that would be go a way to explain the difference -- people not in school, or who intend to flee school, won't be around for literacy screening.
All of this could be avoided if the US simply adopted UNESCO standards for literacy at a given age, but of course, that will never happen.