The flagrant fraudsters are not as dangerous as the subtle ones.
Do you think there are loads of scientists out there cutting and pasting different lanes onto their Western blots or cropping off incriminating bands, or making up numbers on tables or showing pictures of control cells for the 'treated' cells because the treated ones just don't look as pretty as Nature or Science would demand? Who aren't necessarily giving a false picture but just making it look 'cleaner'?
I always suspected there were, for precisely the reasons you outline, Quorlox. But are they really more 'dangerous' than someone like Hwang WS, who spawned dozens of international projects and who got up the hopes of funding bodies (who will now inevitably be more cynical)? It's an interesting question. I think it's really awful, this fine fiddling, and probably fairly common, but whether it's more dangerous than the full-on stuff...not sure. In some ways perhaps you are right, if the whole fabric of scientific integrity begins to erode.