The other thing that is confusing people is that composite scores are not "averages" of the scores that went into them. Composites are made by taking the standard scores of the subtests, adding them together, and then treating that as a new raw score which is compared to the raw scores of everyone else in the norming sample. A new standard score and percentile are determined from that.
In general, when all of the scores that go into a composite are in the same direction from the mean (for a gifted kid, typically all above the mean), the composite is *further* from the mean in that direction. That is, if you have subscores of 125, 130, and 135, the composite might very well be 138. Remember that standard scores are measures of "unusualness." It is more unusual to be really good at a lot of things all at the same time than it is to be good at a few things.
With the percentiles, you're seeing, as y'all have noticed, serious ceiling effects, where the item difficulty range was insufficient. Again, the percentile of a composite is not the average of the percentiles of the subtests -- it's the percentile as compared to everyone else's composite "raw score," the same one used to calculate the standard score of the composite. You can also see ceiling effects in standard score norming tables, but you have to be actually looking at the norming tables to know about them.
