First, welcome!
It should go without saying that it's difficult to say for sure whether this formulation is accurate, since we don't have the benefit of knowing your DC in person, observing his test behavior, subtest scores, error analysis, etc.
That being said:
I can speculate that the evaluator may have felt the VSI was a low estimate of his ability because processing speed was a relative weakness, and both VSI subtests are timed. Possibly he saw an unusually high number of correct responses that were uncredited because of missing time limits. FRI is less impacted by timing; it is also more plausible that it is genuinely a relative weakness, since you report that your child has consistently been just a bit above average in math, while performing at much higher levels in reading, and this is the area that typically is most predictive of math achievement. If what I've thrown out there turns out to be the case, I would still consider this a nonverbal learning disability, although not, perhaps, quite a classical one. Some children with more subtle NVLDs don't start to struggle with mathematics until they reach higher level math, often the spatial parts of geometry (as distinguished from the logical proof part). In fifth grade, even GT math is fairly concrete, and essentially all some variant of arithmetic, which one could argue is actually as much a language task as it is anything else.
The GAI is easily explained. Five subtests go into the GAI (minus the working memory and processing speed subtests): the two core verbal subtests, the two core fluid reasoning subtests, and one visual spatial subtest. If the index scores are representative of the subtest scores, then the GAI was derived from three Average scores and two Superior scores, which is pretty much the same balance as the FSIQ. A simplification: say the VCI subtests average 14, the VSI subtests average 10, the FRI subtests average 10.5, WMI averages 12, and PSI averages 10. The mean subtest score for the FSIQ would be 11.6, while the mean subtest score for the GAI would be 11.8, a negligible difference. The bottom line is that three of his weakest subtest scores are in both the FSIQ and the GAI, and one of his moderately stronger scores is omitted in the GAI (namely, working memory). GAI is not a magic bullet for highlighting all 2e learners.
HIs history and classroom presentation, as you report them, are actually quite consistent with his current testing (with the possible exception of artifactual lowering of visual spatial measures because of low processing speed--but only your evaluator has access to that data).
Psychostimulant meds are a discussion for you and your pediatrician. Anecdotally, I think I've seen them help with processing speed when internal distractibility was the reason for the slow speed. If it's actual processing time that's the issue, they may or may not be helpful. Sometimes they can hurt, as slow processing speed could also reflect deficits in cognitive shift/flexibility (more likely with this profile than with some others, btw, given the relatively lower FRI), and some kids hyper focus and get stuck when on psychostimulants, which makes them even slower. At this point, you don't have a diagnosis of ADHD, so I'd be hesitant to go that route without further evaluative data from someone qualified in that area (like a good child psychiatrist or neuropsychologist--not a run-of-the-mill pedi or other PCP).
