Six variables moderate every omega-3 result: which molecule, what dose, which chemical form, who the patient is, what the comparator was, and what else the patient is taking. Trials that look contradictory almost always differ on two or more of them. The three above differ on molecule, dose, population, and placebo - which is why they can all be true at once.
The practical translation: when someone cites "a study" about omega-3, the first question is never "was it positive?" It's "which configuration did it test?" A null on 1 g/day of mixed oil in healthy adults says nothing about 4 g/day of pure EPA in statin-treated cardiac patients - and vice versa.