Although I dismissed it earlier, your post got me to thinking...
This is actually representative of how simple it is to skew results in your favor.
Let's say that a drug company wanted to hide the fact that there is a correlation between their particular drug and a syndrome like Autism.
Let's also say that, because the've been watching this link for longer than most of us peions, they've got additional data that we don't have; stuff like: autism clusters, amounts of vaccine sales in a correlating area, etc. Let's say there is also a genetic or environmental correlation to autism.
All the leader of the study would have to do is say something like: OK, let's base our study here and, just to discount any unfairness, let's exclude anyone with a history of autism in their family...
While the study would still be "officially" sound, it would tend to point away from their product if their with other causal factors that, taken together with their product, leads to autism.
Not saying it's happening here-- I'm just pointing out that, given the right information and incentive, a study could be designed to clear whomever needed to be cleared.
Who's got the most to lose if a causal relationship is found? ...and, of course, that sword cuts both ways. Given the incentives, though, who's research are you going to believe?
Of course data can be manipulated. Organizations do it all of the time-- when ever they want to be able to give some good news about themselves. Sometimes its blatant, sometimes it's not. In my original post, I wasn't trying to make any grand point--hence, my "small sample" comment-- except that there's more to the problem than mercury as a preservative-- if mercury is the problem.