People with prosopagnosia tend to develop compensation strategies that rely on attributes other than face. Hairstyle, body proportions, style of dress, posture, gait, mannerisms, voice, etc.. Under favorable conditions they (we? I scored 2) can be very good at recognizing people. That's why the test removes everything other than the face itself.

> There's the bizarre phenomenon, which people other than me have reported, how some people seem to consistently resemble photos of themselves, while others don't so much: they're almost a different person in photos.

I'm also just speculating wildly, but this might have something to do with (monoscopic) photography removing depth cues, which is also a factor in the phenomenon of "the camera adds ten pounds". Different lighting and some approaches to makeup and clothing can exaggerate or diminish this effect.