The onus is on you to provide any sort of proof to contradict what everyone else in the community has used as basis of theorycraft. And simcraft certainly doesn't model crit like you're suggesting, as was explained by a previous poster, so your theory cannot explain simcraft results. What can explain simcraft results is that he had to drop a few ilevels and a lot of crit to obtain 4 piece, and that simcraft does not by default include skull banner, making the 4 piece appear a smaller bonus than it is. You've provided no evidence whatsoever from logs that the system works as you claim, and if you go hit a boss level dummy for an hour and look at the log you will see that your theory is pretty much nonsense. The melee crit rate will be close to character sheet -3% (if you use no crit trinkets), special crit rate will be higher (assuming reck on cd, and stopping log just before reck comes available again so that it has maximum chance to be "negated"). Unfortunately even an hour is a fairly small sample so you'll probably dismiss it as "the crit table spreading out in a favourable way" or some similar hand-waving rubbish.
And actually, if your theory was accurate, that would mean auto-attack crit rate should always be below what's predicted because auto-attacks never get the benefit of recklessness but always get the "negation" effect. Even in this example, with predicted auto-attack crit rate of 26.5%, maybe 27% accounting for a small amount of sub 93 mobs (but as a lot of that is aoe that affects a lot less than the full 29% of melee attacks), the actual auto-attack crit rate is bang on the predicted.