I tried and I still see the same: You think that you have 100 combos and 20 of them are bluffs then you have 20% bluffs, and this is the fist part where your thinking gets flawed IMO. There is no such thing as a pure bluff. If you have 10/100 bluff:total combos with 30% equity when bluffing you still only have 7 % bluffs in your range, despite that 10% are bluffs measured in #of combos. The same is true for the range you asssigned to MP. It is way too strong because it bluffs with too much equity. Dont measure bluffs in #of combos, rather in equity vs the worst calling hand in your opponents range.
As I see most people think that the goal of bluffing is to make your opponent fold, but it isnt so, the purpose of bluffing is to make your opponent indifferent between his options for calling and folding so that you force him to pay off your value range. If your opponent can overfold at some point you are not bluffing enough, what ever # of bluff combos you may have.
Also to your assumptions about your opponents ranges: after a few hundred hands you cant make such reads on which you can rely in decisions that need the least amount of precision. What do you know about your opponents turn c-bet tendencies? Does he check with nut hands on the flop or on the turn ? Does he 2 barrel his entire range or just bluffs ? Does he donk x/r ? Is the fish a peelmonkey or a fit or fold guy ? Does he slowplay ? If yes does he slowplay on dry boards and in HU pot or in MW ? IP/OP ? Does he delay until the river or he steps on the gas pedal on the turn when he has something ? There are players with 0.5-1 AF who barrel 99/99/99 with initiative no matter what, there are who c-bet 30% but its still bluffheavy, there are players with 80% turn c-bet who dont bluff enough because they bet with a condensed range, and you just simply cant figure these things out in a few hundred hands. There is a chance for each option after 100 hands, even if you observe them 1 by 1.
If you fold preemtively its your decision but you wont convince me about the correctness of it without real evidence, and for that you need at least 50-100 hands sample size from similar situations (i.e. turn c-bet multiway OOP in a tight range situation)