Inputting errors are indeed common in any dataset. However a properly rigorous meta-analysis should consider data errors, have defined search terms and then include ALL studies that fit the criteria, regardless of the conclusions of the original research.

But then there are proper, academically sound and unbiased meta-analyses conducted by reputable bodies like the Cochrane Collaboration.

...And then there are govt lackeys who try to do a bit of sleight of hand in order to try and dupe the public (often via the media) that their perspective is supported by "science". We all need to use our intelligence and a sceptically enquiring mind on these occasions.

-I'll get off me lecturing podium now....