Exactly. It's possible there are other features that would have more predictive power, they just haven't figured out how to quantify them. The title of the paper could be, "We're still not very good at describing what makes people attractive."