(1)  For the same elevation, test various positions to see how well the corrections work.  This eliminates elevation influences thus, only the WNIR rotator is at play.  5 - 10 test points, get the statistic of well the correction worked, how close to the reference pixel pointing really is.

(2)  Measure the center of rotation, as initially done, at several elevations.  This will test whether rotation is a function of elevation.  How can that be you ask??  the rotator bearing is being bent around by the elevation axle, it's flexing as a function of elevation.  If it's a factor here or not, who knows. Also, the secondary is tip/tilting as a function of elevation thus, it's vector that it arrives at the WNIR port is potentially variable.  I collimate only at one elevation angle, not all so, I don't know how much error this is but I do know from experience, it will vary with elevation.