You're looking at it in an odd manner.

Anhkahar socketed red has 20 strength, 60 hit and 68 crit over Lanathel. Lanathel socketed orange has 14 strength, 10 hit, 73 ArP and 52 expertise over Ankahar. If you cancel out the redundencies and then assume that a person using Lanathel simply replaces the appropriate number of Fierce gems with Etched gems, it comes out to 50 haste, 68 crit, and 6 strength compared to 73 ArP and 52 expertise.

Using the SM stat weights that's 50 (2.97) + 68 (2.11) + 6(3.11) = 310.64 verse 73 (2.87) + 52 (1.97) = 311.95.

Lanathel barely edges out, but it does win, even as an Engineer. You really could use either though. 1 AP isn't even 1 Dps. I would personally prefer Lanathel for the added expertise, myself.

As far as haste being above strength, no numbers I ran - even with absolute BiS/orc racials/ideal professions/etc - could get haste over str. The gap became incredibly small, sure, but none showed it going over, no matter how much tweaking I did. Even if I did mess up, which I doubt, and you're numbers are correct, which I also doubt, and you ignore the multi-target aspects it still wouldn't be worth gemming all haste due to the mentioned scaling issues. EP simulations add the stat to calculate the value. It doesn't subtract it. So, yes, 100 extra strength might not be quite as strong as 100 extra haste, if you're numbers are to be believed. But trading 100 of your current strength for 100 of your current haste isn't the same thing, and the strength is going to be more valuable due to how it impacts every other stat.

That's a large part of the reason why one can only get so specific with EP numbers. Aside from the obvious fact that it's impossible to account for every single factor (many of which may vary from person to person, raid to raid, and so forth), it's also impossible to get one single value (even if you could account for all those previous values) and be able to apply it equally in every situation. Trading stats simply isn't the same as gaining stats.

This is especially problematic when trying to determine a BiS set. You need stat weights in the BiS set to generate a BiS set, but to generate those stat weights you need to already know the BiS set. It's completely circular. In the end, you have to make certain assumptions, and then just accept that things might not be 100% accurate, but the difference (if there is any) between that and reality is incredibly small, and that it's the best you can come up with, so you'll run with it.

I could go on and on - and probably have more than I should, so I'll be quick - but it's one of those things which is very important to understand if you are trying to min-max; you can try you best to do so, but you can never entirely do it (not with being able to prove that you've entirely optimized your gear/gems/etc). Those necks are a perfect example; when two items are a 1 AP difference there are so many assumptions along the line which could push one or the other on top. In the end, simply accept that it honestly doesn't make any difference which is better, that you're never going to know which is completely better, and that likely there isn't one which is better than the other in every situation.

It's the reason why the threads here on EJ are never-ending, even when you ignore patch changes and the like.

Anyways, rambling aside, when in doubt run two simulations, changing nothing but the gemming or the item in question. It's one's best bet in such cases - be it when comparing those two necks or comparing those two gemming methods.

Not sure about the others but are you sure you have 5/5 Necrosis in your testing build Baldless? The damage done by it in your sim is way too low.

This is the problem statement, proc hits much harder than 2k on average. As a matter of fact looking at my festergut logs from last week (when i didn't get tricks) it hit for 3.3k on average which is consistent with the assumption of it scaling with most of our passive damage multipliers/debuffs.

My own spreadsheet shows proc canceling providing me with a tiny boost of ~15 dps on a tank-and-spank fight. Any haste effects though increase the dps gain (lust, engineering gloves, haste pot, etc). Depending on the duration of the fight (and hence uptime on said buffs), amount of movement or aoe (which favors keeping the proc for more spell damage), etc it may or may not be worth it to cancel the proc. I macro'd cancelaura together with melee strikes (and pestilence) to get the uptime of chaos bane up for aoe or out of melee range situations while still canceling the buff fast "enough" for pure tank and spanks.

On a side note, and I dunno if this has already been mentioned, but a great way to compare items with hit to items without it is to set the EP value of hit to be the same as haste when comparing 2 items. That way you can quickly tell whether it's worth it to get an item with hit or swap out some str/haste for str/hit gems. The main problem with this approach is you can "run out of yellow slots" which you can use to gain hit from gems. Still, it's easy to see which items are the "best" choice for hit (t10 gloves, ashen verdict rep ring, and polar bear bracers with LDW's necklace and garrosh's rage trailing behind).