As to digit span, there are exercises that can improve digit span per se, but not the underlying working memory skills, which means they're not really useful. For example, there is a commercially-available product out there that reports data on improving digit span, but the actual research finds that it literally improves -only- digit span. Not letter span, or picture span, or spatial span, or any other kind of working memory task. With limited transfer and generalization to any other kind of cognitive skill, it's something that is unlikely to do harm, but given the time and money one would have to expend, it is also not something I would consider a good return on investment.
In any case, I wouldn't give much additional thought to a DS score in this range (it's still above average), unless you are actually seeing functional concerns somewhere.
The percentiles on most cognitive assessments, including the WISC-V, are based on standardization samples designed to be representative of the population. After some statistical smoothing, these percentiles represent the ordinal performance of the specific student against the general population of their peers (age-peers, in this case). So a 98th %ile means that this student performed in the top two out of a hundred of comparison peers their age on that task.
Subtest scores (such as the ones you've posted) are derived directly, the way I just described. Composites scores (such as the GAI or FSIQ), are derived from the sum of scaled scores that contribute to them, which is then converted to standard scores and percentiles.
