If I were Google, and I'm not, I'd set up those images/labels so that there would be a retention of all the guesses, despite some of them not being a match. Then I'd run the image again at another time, officially prohibit the use of the already matched guesses, and see what happens. That would recognize the fact that one team member sees very differently, but not necessarily incorrectly, and opens up the chance for further description matches.
_________________________
A smile can be infectious. Let's hope they never find a cure.