First of all, I don't find it any less toxic if the troll is on the enemy team, it's just as disappointing to win a 4v5 as it is to lose one, imo.
That being said, your math has 2 flaws. One, your math assumes there is a troll in every game of the elo range in question. That there is a 9/9 chance of having a troll in the game. This is not the case. From personal observation, the chances of having a troll/fresh 30 are roughly 7/12. Not based on observation though. Regardless, this other variable skews your observations.
Two, you're forgetting the math of the elo system, which causes you to lose more elo for a loss than you would gain for a win, thus counterbalancing the 5/9 win ratio in question. Meaning you have about a 4/9 chance to gain elo in the span of 2 games, one victory one loss. You may have a 5/9 chance to win an individual game, but a corresponding loss makes your elo worse than when you started. This is a large margin, coupled with the unpredictability of "fresh 30's", that causes the "elo hell" phenomenon.
I'm not necessarily advocating elo hell, just stating your explanation, which many others have used as well, doesn't make sense in the elo system.