No 2T6, just 2T4 and badge equipment.
Because 2t6 reduces Mangle to 35 energy, in your case, you would want to increase your limit to 20 energy.
Test 1 (NO powershifting macro): fight time 9min 40sec, 332 melee hits, 269 melee crits (44,8%), 96 mangle hits, 94, mangle crit (49,5%), no misses
Test 2: (WITH powershifting macro): fight time 9min 40sec, 307melee hits, 237 melee crits (43,6%), 100 mangle hits, 31 mangle crit (44,8%), no misses
Test 3: (WITH powershifting macro): fight time 8min 53sec, 327 melee hits, 296 melee crits (47,5%), 119 mangle hits, 85 mangle crit (41,7%), no misses
I don't get the same crit percentage for mangle you do for Test 2. Typo maybe?
It's really hard to tell why your dps would have gone down. There are many factors. Crit rate, average damage, debuffs on target, armour of target, when you get your 2t4 procs and trinket/amulet procs.. did you get lucky with the stacking of procs.. did one proc more than the other.. Getting something like a WWS report would be really helpful because it contains pretty much everything. Although removing your proccable trinkets/amulet would make it easier to tell.
Back to the macros... I was thinking that there is some possible optimisation of the script:
/run local f="Cat Form";f=GetSpellCooldown(f)>0 or [b]UnitMana('player')<XXX and IsUsableSpell(f) and[/b] CancelPlayerBuff(f) /stopmacro [form] /cast Cat Form
I've bolded the changes.
This should mean that you can set the energy limits more precisely. I've sometimes had strange energy gains that mean I have 39 energy when I was expecting 40 energy after an energy tick. I noticed the mangle energy cutoff was set to >14.. so if I had had that strange fractional energy gain so my energy was really 14.5, I would not powershift and I would also be unable to mangle next tick.
The second change was just shortening "not IsUsableSpell(f) or" to "IsUsableSpell(f) and".
EDIT: Although I haven't used FeralToolKit yet, nor am I a Lua programmer... The fractional energy gain problem doesn't seem to be an issue for it.