Yes, but as the source explains, the AFQT has several sub-tests, and most of them are highly correlated to IQ. (A few, like the automotive knowledge test, are uncorrelated with IQ.) They claim that appropriate analysis of results on the sub-tests allows a reasonably accurate measurement of IQ to be made.
EDIT: Re-checking the source, they mention that in 1989 the armed forces rescored the AFQT to make it more g-loaded and repeatable. The first draft of the book used the 1980 scoring, but they redid it all using the 1989 scoring because it was superior.