The problem is that various sciences have very different requirements for what constitutes a valid 'discovery'. e.g. In particle physics if you don't have a 5 sigma result you haven't really found something. You often see incredibly weak data accepted into journals in the soft sciences, see the recent 'chocolate makes you thin' faux study as an example.
I think there is a large problem in soft sciences in that they are very quick to declare knowledge with very limited data supporting it. The various dietary sciences are probably the worst offenders at the moment. As you've said we know very little about nutrition and health, so why are we so quick to make definitive statements about it? That sounds like the opposite of science to me.
I'm sure this will be controversial but I'd rather not classify such 'soft sciences' as science at all. I've seen several complete reversals in soft sciences in my life and think that this actually contributes to the general public's mis-understanding and mis-trust of science as a whole. e.g. "Those scientists can't even figure out what I'm supposed to eat so why would I believe this quantum physics non-sense!"