Handling gamut differences when comparing printers

Given a common test image that has not been optimized to either printer under comparison, what is the best way to handle out-of-gamut colors in the image when compared to the profiles being used in the evaluation?

Adjusting the image to fit each printer's profiled gamut would seem to invalidate having an objective image. However, letting the colors clip to the profile using Relative Colorimetric or perceptual rendering may lead to unwanted colors.

Michael, if you have time, I'd be interested in how you address this in your comparisons.

Thanks,
Kalin