The distance between the butt mounting points should be 10 % of the length of the blank. As mentioned above. A minor deviation from this spec has not resulted in significant errors based on my experience, but if you can get it right do it. It will eliminate a possible question about your process.
For AA have you attached a straight indicator right at the tip of the blank and are you using that as the AA refrerence? See figure 8 at this link https://www.common-cents.info/part1.pdf
Can you provide the model numbers and data from your questionable tests? Please also provide the subjective description of the action of the blanks, (slow, moderate, fast, Xfast, etc).
If you make any errors on the high side of IP it will also increase the AA, so be sure you are not over-estimating the IP with testing errors.
The process was designed based on fly rods, but there is no discrepancy introduced by using it on faster action/more powerful rods. The work done by Hanneman many years ago is all still valid.
He added weight to get the natural frequency into a slower range where it could be counted, but if he had had more modern technology, he certainly would have used it. For example, I have found a way to check the true natural frequency of blanks and rods by using a cell phone/Kindle app and it is very accurate. It allows one to see what adding components (guides/wraps/tiptops) does to natural frequency, which most believe is a good indicator of sensitivity.