No need to crop images. The reason why fitting ‘overflows’ to the sister chromatid is because (i suspect) you are specifying an exaggerated ‘Max. radius’ (Path Manager’s Refine/Fit>Parameters… command). Before performing any sort of fitting operation you should have an idea of the size of the structures being traced. A convenient way to do so is to use the ‘Estimate Radii’ command (gear menu in the Auto-Tracing widget). This will give you something like this:
I.e., at the image resolution the max. expected radius is ~7 pixels which is way below the default 40 pixels.
There should not be a need for repeating fits. I would expect that if you trace using Hessian-based analysis under appropriate settings , a single fit operation would suffice.
As per calculating the distances between the two sister chromatids: If it helps, you could programmatically generate paths between paired nodes. The advantage here is that you would be able to measure and analyze theirs distances directly SNT. There is a template script (Script Editor: Templates>Neuroanatomy>Tracing> (…)) that exemplifies how to generate straight line paths.