I'm not a biologist but I'm not sure that follows. It could be that the experimentally-derived structure is 100% accurate to the actual physical structure but g

I'm not a biologist but I'm not sure that follows. It could be that the experimentally-derived structure is 100% accurate to the actual physical structure but getting 90% of your predicted residues to match that is enough to get an accurate prediction of protein behavior and hence "competitive."