My bet is that you are dealing with differential flexing. Period. It is very difficult to do what you are trying to do -- guide with a guide scope and a Celestron SCT because the mirror locks are not actually locks, they just make the mirror harder to flop. That means that you've minimized flexure but still have it which is why they look pretty good but just a bit off (IMHO of course).
Start taking short exposures and lengthen the exposure until you see that the stars are misshapen. If you're collimation is good, then the stars will be perfectly round at 5 or 10 or 15 seconds. If Do this after a careful focusing run, at the zenith and after the mirror has had time to cool down.
If you have differential flexure then the longer the exposures are the more eccentricity you will find. If you have an optical problem it will show up in short exposure. And, Stelios is correct - you've defocused the star to check collimation but that's a pretty crude test.
If you download a copy of CCD Inspector, I think that there's a 30 day trial period you can get a nice quantitative reading about what's going on including the presence of tilt, collimation error, etc.