Covalent bond has no effect on ligand prediction

[BJHardy]

I am trying to predict the structure of a protein containing a flavin post-translational modification, where FMN is covalently bound to a threonine residue via its phosphate group.

I first tried including this as a modification on the threonine residue using CCD "FMN", however the linkage is incorrectly predicted:

I then tried predicting the FMN as a separate ligand and defining a covalent bond between the threonine oxygen and the FMN phosphate. However, the covalent bond seems to have had no effect on the prediction. As you can see here, it has just made an arbitrary bond.

How can I ensure that the covalent bond will affect the position of the ligand?

[gcorso]

Hi @BJHardy could you send us the configs/input files you used?