There are (or could be, due to unknown factors) a few reasons:
1) I'm not sure what fight length or iterations were used to generate that chart.
2) The talents used in the T19P for Shadow need to be updated. It currently uses ToF/VL/SL/PI/LotV when, according to the comparison of all DPS talents I did in the first wave here for Heroic (which, again, may not be the same at a lower level) [
https://www.beotorch.com/batch.php?r...c-008cfa070490 ] show that the best non-S2M talent combo is ToF/VR/AS/PI/MSp by about 3.3%.
3) Mass Hysteria affects S2M more than other L100s, and at a lower profile level you have fewer other artifact traits to balance this out (or reduce the overall impact).
4) Stat distributions, especially trinkets, are different between the two levels.
5) Running a sim with all actors that have execute (like 28 Shadow Priests) as opposed to a sim with only a few actors with an execute (like the stack rankings for a raid) mean that you'll have more time spent in execute range, which means higher uptimes on abilities used within that part of an encounter. There's not an easy way to correct for this, so I try to do my base sims with as wide of a basecase as possible. One flaw in this set of batches of mine is that I should have always included all 27 combinations of L15/L60/L100 and pivoted the batches off of a different L75+L90 combination to mitigate this as much as possible (I pivoted off of L60+L75 instead, so RoS isn't used in every sim run).
It is worth noting that while, in the chart linked, there is a ~22% DPS difference between S2M and what is likely the wrong talent combination for T19P, the delta between the top S2M and that same talent combination on my simset for Heroic is 10.6%. How much of this is attributed to each of the 5 points above? Impossible to tell, though my money is on #3 for a lot of it.