The 95% confidence interval for a GAI of 110 is 104 to 115. So if you took 20 kids who scored as your son did on the first administration, 19 of them would have a "true" score between 104 and 115, and the other one would be farther out in one direction or the other.
For the second administration, the 95% confidence interval is 132-143. So there's not even any overlap, such that you could say his "true" score would lie in the overlap and the test results were consistent with one another.
If there wasn't some change in your son (unresolved physical or mental health issues, lack of sleep, lack of cooperativeness, etc.) between the first and second administrations, I'd also suspect administration issues.
