I believe that some tested kids wind up not testing as "gifted" as adults in the sense that a child (especially a child younger than 8) who may have tested at say IQ 130-135 can certainly test at 116-121 (less than one standard deviation) on an adult test.
However, don't confuse absolute numbers with percentages. In your example, there can still be 2%, for example 6 out of 300 adults who were formerly five to seven year olds. As for the GAI issue, it is meant to include more people who may otherwise be excluded due a relative weakness in working memory and processing speed; it is possible, even likely, that such individuals as adults may not make the cut if tested on an instrument that didn't remove working memory and processing speed factors.
