When you adjust the collimation screws the star will move out towards an edge of the field of view because the secondary mirror is moving. You then recenter the star. If the star is not well centered then, as I understand it, because the star is now an extended object rather than a point souce and the scondary is convexly curved, the appearance of the diffraction rings can be affected by the curvature of the mirror.
If I'm being really pinickety on a night of imaging, I'll use my ToUcam and the reticle function of K3CCDTools to make sure my defocused star is dead center of the screen. That way I can be pretty confident my diffraction rings are off center because of collimation and not because of where the star is in on the secondary.
Does that make sense
