Welcome!
Before actually computing the GAI or FSIQ, I can tell you off the bat that they are unlikely to be substantially different (though yes, GAI should be slightly higher). There are three important discrepancies in this profile, including the PSI. The VSI and WMI are also much lower than VCI, but the impact of the relative weaknesses is actually highest for the VSI, because the subtest it contributes to the FSIQ and GAI is the lower of the two (Block Design). The other two indices contribute their higher subtest (in this case).
Portia makes a good observation--the lower subtests are largely in tasks affected by visual-spatial skills (VS and PS) (though I would include Picture Span only provisionally, as it isn't significantly weaker than Digit Span). And in answer to how this might relate to a child with strong interests in science: up until he secondary level, when quantitative aspects of science become important, an interest in science is easily supported by strong verbal gifts. There are minimal mathematics demands of even the arithmetic kind, and virtually no sophisticated quantitative reasoning demands. (And clearly, your child's assessed fluid reasoning is more than adequate to handle the ones that do exist at this level--and likely even at higher levels.) Although fluid reasoning is "only" in the MG/HG range, that is certainly quite strong enough for the math required in most STEM fields.
On the changes in achievement scores: the changes in reading and written language are within the confidence interval, and really just mean that he has kept pace with his own historical rate of learning; they don't represent any special gains. Correspondingly, the drop in math score doesn't mean he has lost any skills. These are standard scores, which represent the ordinal standing of an individual in their age-peer group. The change in math likely represents the changes in expectations between primary and intermediate grades. Two years ago, his nationally-representative peer comparison group had not yet been exposed to instruction in multiplication, and most of them were not fluent in simple addition or subtraction. Having even rudimentary multiplication skills would catapult a student to the top of the norms. This time around, the peer group should all have been instructed in all aspects of basic arithmetic, and is expected to have fluency in addition and subtraction. The curve is a bit broader at this age, which means that it takes more to go from above average to stratospheric in the norms.
In any case, these achievement scores are still roughly within the predicted range for this cognitive ability. To your question of how to compute the Total Achievement: that can only be done with the publisher's proprietary webware. You'd have to ask your examiner for that.
Bottom line, I don't see data to support meaningful changes (let alone regression) in normative academic standing. However he performs in the classroom with regard to time pressure, he did not do poorly on the WJ. If you want to see if timing has an impact on math achievement testing, I'd check to see how his Math composite (vs Broad Math) looks, as that doesn't include the timed fluency subtest. Similarly, look at his Academic Fluency composite vs Academic Skills and Academic Applications.
