There are many reasons why percentages are misleading. They can mean that the move is more often played when white is stronger than black, or in situations where a draw isn't what white wants, or there is a trap that many people fall into but that can be avoided, or lots of different reasons. Maybe it used to be considered good in the past until some novelty was found that refutes it, and then people stopped playing it.
For your game, what matters is how you will score playing it. That depends not on the results other people had with it in the past, but on how good your moves in the rest of the game are going to be.
So much more important than the winning percentage is how well you understand the idea behind the move. 5.Nge2 there is a bit odd since it doesn't pressure e5 and blocks the white bishop; but in return you keep options open with your f-pawn (you can still transpose to the Samisch, basically). Is that what you want? Do you think you can make the best of those pros and cons? Do you understand what black can do to make use of the cons, and are you ok with that?
That's the sort of thing you need to be asking instead of looking at percentages.
Although of course if they are extreme and you don't know why, that may be a hint that there is something you need to figure out.
Also, the problem with "is it better to play the less common because your opponent is less likely to be familiar with it" with a move that you are not familiar with, is that you are certain that you will be giving yourself that disadvantage, but there is still a chance that your opponent actually knows the move well. So only do that with sidelines you actually investigated a bit first.
