I get what you're saying, Gusto, and I agree - taken literally (i.e. on the assumption that somewhere there was a sample of children who all took the same tests and percentiles were calculated from them all, no funny weighting) it doesn't make sense. Someone who outscored or equalled your son overall cannot have dropped more than 2 marks in total, so must have outscored or equalled your son in at least one of the two subtests where he dropped a mark (because if someone scores strictly less than him on both of those two, they drop at least 4 marks in total). Therefore, the total number of people who outscore or equal him cannot be larger than the sum of the number of people who outscored or equalled him on those two subtests.
Possibilities other than simple mistakes might include: funny weightings; smoothing/extrapolation; a sample who did not actually all take the same tests. I agree it's weird, and I also agree with all the people who say that this test did not give any kind of ceiling on what your son can do. Welcome!
