Due to random chance, you'll sometimes get positive results, and sometimes get negative results. If all of those studies are published, then there isn't a huge problem -- a metastudy would show that on average cocoa doesn't affect health.

But instead, support that only the positive studies are published. A third party reviewing the relevant literature will assume that all studies conducted showed a positive result, when in fact they were only the studies that were published.

This is the problem with having sponsored research: the research itself can be done properly, but a bias in publishing can give an inaccurate view of the true science at hand.