I don't think they are faked opinions. Sure, there's always one or two fans of other games that like to pay a visit to a competing game just to take a bash at it (probably motivated by the fact that other people did the same to their own favourite game), but for most of it I think people simply set themselves up for disappointment by giving more importance to their own hopes of what the game would be rather than to the mountain of hard evidence pointing in another direction.
You assume that buying without trying a demo first is the same as buying unseen. Trailers and gameplay footage is however widely available through YouTube, Facebook, Instagram and other platforms. Games have their own websiter, articles are published on gaming sites and games are discussed in forums. There are plenty of ways to see a game without actually trying a demo. Furthermore, demos are not available for all games, which skews the statistics.
Risk research has very little relevance. You look for information primarily because you are curious, interested, hyped - not because of some risk.
That is the reason for buying though. If 4% of gamers base their decision to purchase on reviews and advertising, it doesn't mean that the other 96% of them don't read reviews and aren't exposed to the ads. You should also note that the sum of the 2015 chart is 100%, which means that either it measures only the primary factor of influence, or it is a multiple choice question measured as a percentage of all influences instead of a percentage of all gamers, so in either case the influence of ads and reviews is likely much greater than what the chart suggests. Which probably is the reason why they changed the design of this chart between the 2015 and the 2017 documents, where you can see that the influence of graphics went from 7% in 2015 to 67% in 2017, and Online Gameplay went from 4% in 2015 to 50% in 2017. Pretty massive difference, and it means that the 2015 data is essentially useless until we know what they measured and how the percentages are calculated. You certainly can't compare the 1% advertisement influence from 2015 to the 67% graphics influence in 2017, because these percentages are not calculated in the same way.
Your data is about what influenced the decision to purchase - not about what the customers knew when they bought the game.