I did some more poking around, and varying sources indicate that the NARA-III generates standard scores, age/grade equivalents ("Reading Age"), percentiles, and stanines (but many consumers use only the reading age). I'm going to guess that his raw scores were well above the score necessary to obtain the GEs you were given (hence, hit a ceiling), which is why the GE/AE is a considered to be a low estimate. But that would suggest that the max GE/AE obtainable on the instrument is fully two years below the alleged range of the test. Still odd.
Anyway, the results do seem grossly in line with the WJ results.
