1. STAR underwent a substantial structural revision in 2011.
2. Yes, if she has been near the hard ceiling of the test for a while, her percentiles should be falling as more children run up against that ceiling.
3. Sort of: The instructional reading level is supposed to be a criterion referenced measure, that compares her to an absolute standard (80% accuracy in reading comprehension of text at this reading level). And as an adaptive test, it should be giving her the same items a 12th grader would be seeing, rather than indicating the performance of a 12th grader on seventh-grade items, except that after 6 months, it resets to your grade-placement for the start point, so it is possible that she's starting over at just below grade level for the level of the first few items every September. So you'd think that later administrations in the same school year would be closer to her real level...
4. Except that it won't recycle the same items for 90 days, so if she exhausts the hardest items early in the year, she might actually have her score derived from easier items later in the year, because the hardest items aren't allowed to re-enter the pool yet.
5. The GE, which I think is what you report, is a measure distinct from instructional reading level, and does max out at 12.9, but is NOT a criterion-referenced score. It's a norm-referenced-derived score, and means, as GEs usually do, that it's the score that was obtained by the mean student of that nominal grade placement, NOT that she reads at that instructional reading level.
A little light reading, for your perusal:
https://resources.renlearnrp.com/us/manuals/sr/srrptechnicalmanual.pdfSTAR math maxes out at +3 grade levels above nominal grade placement.
http://doc.renlearn.com/KMNet/R0054119FEC4F604.pdf
