The basic issue is that reviews are just looking for glaring problems in the presentation. To really test an article one has to replicate the experiment from the ground up. And these days thats damn hard and expensive to do.
http://www.nature.com/news/1-500-scientists-lift-the-lid-on-...