Given the lack of a sample face, I'd say your result might be nearly acceptable, since there's a bit of similarity between some of the faces presented.
If this is just running a % match off some facial recognition software, then it's completely missing the point of celebrity look-a-likes.