MW optimizing stacked healing

[Teabender]

Raiding 10 man in a 3 night guild.
worldoflogs.com/guilds/8230/ (not allowed to post links)

So, I'll just dive into it..

Mana is pretty much never an issue now as most of the MW community agrees upon, so what I'm trying to maximize is my throughput while grouped up (Thok P1, Malkorok P2 etc.) as that is what is giving us the most problems at the moment.

Now, I understand that Thok's and Nazgrim's are the best for this scenario, but until I get those I'm using Prismatic Prison of Pride with heroic Horridon's.

For the Scenario above, what would be the recommended, to stay at the 9158 or go to the 12138 one (12138 might not be an option when not using AMP trinket I guess)?

It was the first time we killed Thok this week, and I started off with Xuen and Chi Burst (will be going ZS next time). After the 3rd try I just thought I would give RJW a try, and it ended up doing a lot more healing than I could push out with SCK, but probably resulted in less uplifts as well. Overall healing is hard to judge because of RNG and the small sample size, so looking for theory around it.

What I'm wondering is with with 2 piece set bonus (and 4 piece when I get there) and taking trinkets (Thok's & Naz's) into account, which haste breakpoint would you recommend for heavy SCK / RJW healing.

As I understood it mastery spheres proc chance was reduced for SCK, is that the same for RJW? Or does the power of RJW change compared to SCK with higher haste levels (I know haste scaling is the same, but if chance of mastery orbs is different, does this favour RJW with more haste?)

I just ended up really liking the throughput with RJW when grouped up, but if that isn't supported by theory I'll put it off as "lucky accident" and go back to SCK and Xuen again.

[Astraios]

PPP is generally agreed to be one of the best trinkets, then cleave for stack or nazgrim for not stack.

That's what I always read anyway, hard to tell w/o testing, but PPP is very good either way (also encourages gemming secondary stats even more than before.)