Clarification on the application of SfM

[DakotaFrico33]

First of all, congratulations on your work. I have learnt a lot from reading your paper.

While reading the paper, I got confused with the part about SfM. My understanding is that SfM is needed to reconstruct 3D structure of the scene when camera poses are unknown.
My question is this: if the camera poses are already given by ARKit (or ARCore), what is the reason for applying SfM?