The mistake that many people make, and that I think you're implicitly making here too, is this notion of "controlled study". If the study denies the result seen time and again in real life, perhaps the study isn't studying correctly.