Yes, FSIQ and even GAI are not good representations of her ability. 95% actually means the confidence interval for the calculated score, not whether it's a valid representation of her.
Sounds like she may have had a different test, or different selection of subtests, last time. Also, persistent reading decoding weaknesses can drag down oral vocabulary scores over the years, as kids then have less access to the high-level reading vocabulary than a child with that verbal intelligence would otherwise have. The reading disability interferes with the ability to keep pace with their non disabled high-cognitive peers. Remember that this is a norm-referenced test, which compares her to her age peers, not an absolute standard.
The migraine meds very likely do affect her already-slow processing speed. And possibly even her overall test performance.
