Despite what simulation tools say, haste is not a very attractive stat over what we cannot avoid from gear. "Faux reforging" over several valus of secondary stats between haste, crit and mastery show that;
mastery and crit have a very high synergy
haste and crit have an okay synergy, though not as good as the one between mastery and crit
haste and mastery diminish eachother - in all cases it would be better to go either full haste or full mastery, with full mastery being significantly better within the observed min/max
In addition I also found that
mastery and crit are very close together in value
mastery will be better then crit at more intervals than vice versa
a balance of mastery and crit shows better results than going purely for one or the other
while haste may show slightly better results at specific increases, it also does so at specific decreases, and shows loss just before and just after these points - this can only be due to dead time or shot shifting.
haste will, apart from the specific shot shifting points, be a loss over mastery or crit at all other tested points
This leads to a valuation of:
Agi>hit==exp (to cap)>>mastery>~crit>>haste