Rosetta structure refinement alters bound ligand

I'm refining tubulin based on this tutorial here: https://www.rosettacommons.org/demos/latest/public/electron_density_structure_refinement/structure_refinement

I save the nucleotides and separate pdb files, convert them into mol2 and generate params file that I pass to Rosetta.

It seems, however, that Rosetta completely flips the orientation of the nucleotides. 

I attached screenshots of both the starting model nucleotides and the refined nucleotides fitted into the density.

On a side note, what article is this atomic refining based on? I want to learn more about how the structures are refined into density. 

Regards.