It makes a lot of sense to compare you to other people playing the same thing in the same place. How could you grade a mid zyra against a support zyra? The goals of support are completely different.
Yes, there are ways to abuse the system (like taking raka and building adc) but overall it works well for people playing champs where they're supposed to go.
As far as what the performance categories are, that is undisclosed. Theoretically, any statistically significant number from the game could matter, such as damage dealt, CS, gold, kda, wards placed, game time, etc. I assume that the specific equation or weights of these various numbers differ based on role (ie adc gets more weight on kills than assists, supports get more weight on assists and gold) but that is speculation. Once the algorithm has spit out a score, it compares it to all the other scores from players who have taken the same champ in the same role, and assigns grades based on what percentile you fall into.
In other words, if you take a champion somewhere strange (Cass jungle) and do ok, you stand a higher chance of being in a higher percentile, since there are fewer people who have played that combination and done well.
In your specific case, you did above average when compared to other ekko junglers, but poppy did amazing when compared to other poppy tops.
