I believe that the win ratios being this way are due to the possibility of mirror matches. Le'ts say we have a fighting game with 2 characters: A and B. If this game does not allow mirror matches, then win percentages between these two characters would always add up to 100%, because when one character wins, the other always loses.
For example, A fights B 10 times and A wins 6. This means that A has a win percentage of 60 and B has a win percentage of 40. However, things get tricky when you add into the mix the possibility of mirror matches. Consider the following scenario:
A fights B 5 times. A wins 2 times, meaning B wins 3 times. A fights A 5 times, meaning A always wins (A always loses as well). Overall this gives A a win percentage of 70 (7/10) and gives B a win percentage of 60 (3/5). As a result, even though there are only 2 characters in the game, they both have a winning percentage. Assuming that NRS is using an algorithm as simple as what I described to calculate win percentages, then these statistics are possible.
Truly, there is no way of knowing whether this is actually possible without knowing the algorithm that NRS is using to calculate win percentages. These are just my thoughts and speculations. Let me know if I've made any mistakes, or if you have any alternate theories to share.