different schmidt collimation inside and outside of focus

I was carefully collimating a new Celestron 9.25 with a
celestron .63 reducer last night and I noticed something odd. I was
using robofocus to move inside or outside focus the same amount
(perhaps 10 wavelengths). When I collimated inside to be concentric
rings then went outside it was perhaps 10-15% off. When I
collimated outside and went inside it was 10-15% off. I ended up
collimating in the middle to balance the two figuring whatever
effect it was would cancel at focus. I used focus max to focus and
the resulting stars were about 2.6 arcseconds FWHM in a .06 second
exposure using the focus function in maxim. The stars seemed symmetric.

My question is, what optical configuration problem leads to the
optimal collimation being different inside and outside. I could not
find anything in Harold Suiter's book to explain it, and I don't think my
secondary is 10-15% off in the center of the corrector

Thanks

Web