I just tested Composure and Seigan/Hasso, and they're separate terms.

Code:
Stoneskin, 77 RDM/SAM : 26 second recast
Stoneskin, 77 RDM/SAM with Hasso or Seigan: 39 second recast
Stoneskin, Composure alone : 33 second recast
Stoneskin, 77 RDM/SAM with Composure and Hasso or Seigan: 45 second recast
When Byrth is done determining DA/TA values, I can take him out and do the same test with Impact I guess. Scholar JAs (Strategems and the Arts themselves) are another ball of wax, too.

I'm trying to consolidate the information onto this page, see anything wrong?:
https://www.bg-wiki.com/bg/Spell_Recast

Actually, I think all my information about forcing 10 bit operations is wrong. Everything we've found indicates that it's one big floating point equation with a low-bit output, right? Why do they force their variables to be low bit if they're not going to use that to make server-side calculations more efficient?