Question Question about Retarget-from / Retarget-to split advantages

Hello,
I try to explain here what I understood about the 2-phases split advantages performed by Unity when importing external humanoid FBX animations:

1 - Import an FBX with a new Avatart in T-pose, from which I'll extract the Avatar definition. When I specify "humanoid rig", a Muscle clip configuration is created: for example, Unity specifies that avatar shoulder bone rotation range is [-90,90] degrees.
2 - Import an FBX with a humanoid "rise arm" animation. I want to apply this specify that the avatar definition must be retrived from the avatar created in step 1. Retaget-from phase is executed: "Convert a standard skeleton transforms animation to a normalized humanoid animation clip (muscle clip)". This means that if in my "rise arm" animation my shoulder bone rotation range is [-45,45] degrees, they are converted in normal space, depending on shoulder Muscle range specified in step 1: in this case, the new range becomes [-0.5,0.5]
3 - When "rise arm" animation is performed on the Avatar, to know the effective bone rotation, Unity at runtime performs the Retarget-to phase: if in the animation clip we have that at frame N the shoulder rotation is 0.5, the final shoulder rotation at that frame will be 0.5 * 90 = 45 degrees.

I'd like to know:
* Is this workflow explanation correct?
* Who decide the Muscle clip configuration at step 1? Is this information (the rotation bones extension) encoded in the FBX file from the 3D animator who performed the rigging/skinning phase?

Thank you!