Welcome!
Some context may help:
1. cognitive (IQ) scores are not considered really stable until children are a little bit older, often around age 8 or 9, and some children with neuroatypicalities (such as ADHD) don't really settle into being fully testable (i.e., showing the full range of their abilities in on-demand testing) until even later than that.
2. when considering whether the difference in scores is relevant, we typically consider whether it is statistically significant (often p<.05 or .01), rare (base rates in the standardization samples often <10%) and meaningful (real-life impact). It's important also to recognize that regression to the mean has much more dramatic effects with outlying scores.
To give you a sense of the ballpark for significance, it may help to know that most achievement measures have correlation of about .6 with most cognitive measures. So, for example, if we took his FSIQ as the measure of cognition, and tried to predict achievement, we would expect achievement scores around the 75th %ile. Conversely, if we start from the achievement measures (let's say 98th %ile, since that's where they cluster), and try to estimate cognition, we would predict cognition at about the 90th %ile. The latter is actually pretty much where his scores fall.
2b. here's another way to ballpark this: if the 95% confidence intervals for the measures you are comparing touch/overlap, then a reasonable rule of thumb says the difference is not particularly significant. If there's a big gap between the envelopes, the difference might be more worthwhile investigating.
Bottom line: the scores are neither necessarily fixed/best representation of ability, nor are they necessarily "wrong" per se. He is very young, has never (I gather) experienced this type of testing before, and has neuroatypicalities. I would take these scores as a reference point in time, and no more.
