GAI consists of both VCI, both FRI, and one VSI subtest. Since your DC's FRI was the second lowest index, and both subtests were included, a relatively low area comprises 40% of the GAI. And if one of the VSI subtests was higher than the other, but the low subtest was the one included in the GAI, then the GAI could easily be lower than the FSIQ. Likewise, only one WMI and one PSI are included in the FSIQ, and if it happened to be the higher subtest in each case, that could also cause the FSIQ to be higher than the GAI (which doesn't include any WMI or PSI subtests).
If you're comfortable posting the subtest scores, I could probably explain this a bit more specifically, but this is the general outline of how this kind of disparity occurs.
