I believe that the publisher of the CogAT recommends testing out of level (i.e. using a test designed for a higher grade level) when looking for high ability. Our schools don't do that but it isn't a bad idea b/c it raises the ceiling such that a few missed questions don't mean the difference btwn a really high score and a much lower score. For instance, when you are testing kids who are expected to be near as high as the test goes, they need to be virtually flawless (missing no questions) to get a 99th percentile score. On a higher level test, a missed question or two or three won't drop your score from the 99th to the 90th b/c it is expected that even really able kids won't know everything on the higher level test.
The norms you got should be as compared to 2nd graders who took the 3rd grade test locally and nationally and the age percentile should be as compared to kids his age who took the 3rd grade test.
