That's true to an extent. Especially when assessing defenders. They award more aggressive defenders than one's who are more conservative for example. In regard to attacking players, their rating is pretty accurate ( by attacking i am also referring to central and attacking midfielders). Even if their rating is slightly off during one particular game, the consistency of keeping these ratings means that they usually have it right by the end of the season.
As to your point regarding player comparisons, you are right. For example, players on teams that win more will have better ratings as the team as the team as a whole would be able to achieve the statistical benchmark to receive higher ratings. Again, sense has to be applied when making assessments. A midfielder for sheffield united is far less likely to have a higher passing percentage than a player in the same role for Man City. A player like Phil Foden for example, if given regular game time, might achieve a mid to high average rating like 7.2 for City, but if playing for Sheffield United where they would lose more games and have less control of games, and in which the passing options and dribbling space available for City no longer exist, he would recieve a lower average rating as he would fail to reach the required statistical benchmarks.