The original purpose for journals like PLoS ONE was to give scientists an avenue to publish replication studies. To make doing that have more impact. However I don't see too many of those studies coming out of PLoS ONE. This is indeed a problem and one where we should collectively be going back and figuring out what is fact and what was a function of poor experimental design. It's difficult to make progress when you are basing your experimental design on flawed data. Any time I have come across mistakes in the publication chain I have corrected what I found with in one of my papers. Aka publiched something and repeated a prior experiment set (which was relevant) and reported what difference I saw compared to what was previously available