The question isn't "Do reviews affect sales" rather the question is "By how much do reviews affect sales?" If even one person is motivated to buy a game they wouldn't have purchased by a review then we've established a causal relationship (I can vouch for that). However, the existence of highly reviewed but low selling games suggests that the degree to which a review benefits (or hinders) sales is low.
Of course, then we need to take into account how many people read the reviews. If a game is highly visible than more people will read the reviews and base decisions on them (which can go either way). If a game is less visible then fewer people will seek out reviews for it and those reviews will have a smaller overall impact.
In the end, I think there are too many variables to account for to make a decisive conclusion.