Ok, this one is for Hidden and Chromean. Actually, this is for everyone who uses the EP calculation system.
I did some testing with Chromeans config and I came to conclusion that the sim is working right. The problem lies in the big variance it is giving and that is the result of all random procs, trinkets, flurry, elemental devastation etc. As everyone knows, 2 random procs can give very different results depending on do they happen at the same time or different times, even if the uptimes are exactly the same. If they happen at the same time, resulting DPS is higher than if they happend at different times. And as the number of random procs go up, the variance increase is not linear, it is more like exponential.
As the DPS has increased a lot, you should also increase the EP ranges. Now that is a bad thing, because I think the optimal EP ranges are an amount, which one item can increase one stat. This method should give quite accurate EP values if calculated after every item change. Anyway, I had to up the EP ranges to these values to get some meaningful results:
ep_ap 300
ep_crit_rating 100
ep_hit_rating 140
ep_expertise 5 //not rating
ep_haste_rating 200
ep_armor_ignore 600
ep_spellpower 300
ep_agility 120
ep_intelligence 200
As you see, the expertise is still low. Otherwise it would go over the cap and give weird results. Also the hit rating is questionable as with Chromeans config, spell hit will hit the cap sooner than melee hit. I was trying to get the total DPS difference to match about 180, which is 5% from the baseline total DPS.
Also the 5000 simulation time is not nearly enough for EP value calculations with these number of random procs. I honestly don't know what is enough, it kind of depends on the config I think. In test, I used 10000 hours and EP value difference between 2 tests was maximum of 0.12. Except with expertise(got 2.07 and 2.79), which was expected with too low EP range.
So the advice is to up the simulation time and up the EP ranges, but watch the expertise and hit cap.