What analysis are you using?  For several analysis, a paraxial image height solve is used on the intermediate surface to ensure that those analysis which are designed to be used “at focus” are at focus.  If you choose the actual image surface, the assumption is your system is already at focus.  So you might want to try to place an image height solve on the thickness of the last surface to see if this works.

This is a Double Pass. Image has 0.0509 waves (line 31), line 30 has 630.69 waves and all power. If I tilt more than 0.06deg (line 4) Zemax gives an error message. (Cannot calculate Zemax data for 100% vignetting.

@Rich_G

I took a quick look at your model, and it seems that the wavefront error analysis is working as explained in the user documentation.  Specifically, as noted above by Michael, when the evaluation surface is not the final image surface, then a paraxial marginal ray solve is applied to the evaluation surface.  In the help documentation, see

The Analyze Tab (sequential ui mode) » Image Quality Group » Aberrations (Image Quality Group) » Ray Aberration (Aberrations)

In your case, when this is done, the evaluation surface thickness increases significantly:

Your optical path, which appears to contain Wolter surfaces, is rather complicated, so it’s not surprising that a paraxial ray solve yields this kind of a result. 

I’ve simulated Wolter grazing incidence telescopes in the past and have found that a mixed mode model works well, in which the Wolter surfaces are implemented within a non-sequential surface embedded in an otherwise sequential system.

Regards,

Jeff