But in the study, there is no rating by others, simply by yourself. So if you really wanted to make a statement about people actually being more or less beautiful than they themselves think they are, you would need a way for someone to assess themselves as if they were a stranger. I know a few times when I looked at a mirror and didn't immediately process that it was a mirror, I saw myself "objectively". I wonder if there's some way you could incorporate that phenomenon into a study. I would also suggest amnesia patients who have forgotten what they look like but 1) I don't know if that happens 2) people with amnesia can still remember things in a certain way (like how to play the piano), so I don't know if that would affect their own rating as well.
People studies are hard because there are just so many factors that affect the results, and it's not always intuitive what affects what. Really, most of these studies only directly support exactly what they tested, in this case: people choose the most attractive version of themselves when presented with different variations.