"How can most of anything be above average?"
Five is only average if you assume a symmetric distribution. Getting a very low score requires extreme incompetence on the part of the developer, so there shouldn't be many games actually released in that state in the first place - any game that deserves a 2 or 3 should be cancelled, not released. But you still need a place on the scale for the games that do get released in such a state. A sensible rating system, and one I think most people actually use even if they don't consciously think about it, would look something like this:
1) Broken and unplayable.
2) Will just about run, but that's the best that can be said for it.
3) Extremely buggy, poorly designed.
4) Still bad in most ways, but some people might actually play all the way through.
5) Playable, but not really well made or fun for most.
6) Has some good points, but too buggy/badly written/etc. to be well recommended. Probably enjoyed by people who are fans of the genre or series.
7) Not great, but good enough to be enjoyed by many. Most competent but generic games will fall here, as will innovative but flawed ones.
8) Really quite good.
9) So well made that even people who aren't usually into this sort of game might enjoy it.
10) Fucking amazing.
Obviously on a scale like this, actual finished games by professional developers are unlikely to score below a 6 or so at worst. But if you therefore decide to remove anything lower than that and adjust the scale so 6 becomes the new 1, how would you distinguish between a mediocre game and a truly terrible one? Essentially, you don't want a scale normalised to the average quality of released games, but one in which all possible qualities have a place regardless of how rare they actually are in practice. No game should ever be released in a state where it can score a 1, but it's possible for it to happen so any review scale needs to allow for that possibility.
On top of that, there's an additional sampling bias. People don't play or review all games. The region below 6/10 is not well represented because very few people are stupid enough to buy the games that would fall there, and very few reviewers waste their time on obvious crap that no-one is going to care about. Games that actually get reviewed are likely to score above average because that's a deliberately biased sample of the set of all games.