The article "Fatigue tests of piping components" by A.R.C. Markl said:
"the section modulus Z used in equation [2] is that of pipe of the size and schedule or weight with which the fitting is intended to be used, wich is not necessary that of the fitting itself. the difference may be considerable as in the case of forged tees where excess metal is distributed variously to provide reinforcement for the branch opening."
This phrase is impossible to understand in respect to bend "section modulus Z used in equation [2] is that of pipe of the size and schedule or weight with which the fitting is intended to be used". We could use bends with various wall thickness. It's not the same as the pipe. And It could be different at various manufacturers!
The SIF for bends i=0.9/h^(2/3) is not obtained from experiment, It obtained from theory (Clark&Reisner, Hovard, Wigness, Kostovetsky). Markl's experiments just confirmed the theory. So I think It's better to use section module Z for bend, not Z for matching pipe...
But there's no theoretical data for tees, so we have no other choose except to use Z for matching pipe...
In Russian codes we use Z for bend itself, not for matching pipe