Discrepancy between image w/ embedded and actual profile

I’m trying to determine if an image with a press profile assigned to it will fit within the gamut of my proofing device. I’ve already compared the press profile to my printer profile, and confirmed that a couple of areas in the press profile aren’t completely covered by the printer/proofer profile. OK, fine. What I want to find out next is whether or not the image I’m wanting to proof actually contains any pixels in those out of gamut areas.

What I found was that when I brought my image (with press profile embedded) into the 3D gamut viewer of ColorThink Pro and compared it to the gamut of the exact same press profile, they didn’t match up. It seems as though the embedded profile is being recognized by ColorThink Pro, since there is a difference when comparing the embedded version of the image to an untagged version of the image. What I don’t understand is why there appears to be such a difference between the embedded image file and the profile itself.

Doesn’t the act of embedding a profile restrict the pixels of the file to said profile? If so, I would expect the mapped pixels to all fall within the gamut of the same profile, correct? I’m finding I can’t trust the results of my embedded image (press gamut) vs printer profile (proofer gamut) at the moment… Any ideas?