This is a really great observation. A lot of the time when people say "but that problem is NP!" is it still likely that you'd be able to solve it by exploiting some structural properties of the specifics.