Since you are familiar with regression to the mean, it may also make sense to you that the more extreme a score is to begin with, the greater the regression it is likely to experience on retest. This doesn't mean that it will always do so, of course.
Another factor involved here is that older students as a population are closer to the absolute measurement ceiling of the test, which means that there is less "wiggle room" for small careless errors to occur without impacting normative scores.
Quantitative analysis for you on regression to the mean: special group studies cited in the technical manual note that a group of identified GT students who all had previously been tested in the 130+ range had a mean FSIQ on the WISC-V of 127.5, with a standard deviation of nearly 9. So that means that of the group whose -lowest- score coming into the study was no lower than 130, half of them scored 127 or below (i.e., fewer than half of them scored 130+), with 16% scoring lower than 119. The publishers did not consider this to be evidence that the new test was flawed, but actually confirmatory evidence that it was good.
Without any other information, it is unclear exactly how much is owing to regression to the mean in this case, but it is not out of the range of what I and others have observed in learners around this age. Keep in mind that it doesn't mean that either score is "wrong".
