[phenixbb] Geometry problem with the ligand

Hello all I am working with a 3.0 A dataset, which binds with cAMP. I have refined the dataset with phenix (TLS+Group_adp) and followed by weight optimization. Though cAMP fit well in the density, validation tools (molprobity, procheck) suggest the bond angles and bond lengths of ligand of completely refined model is very bad(100% bad). Can anyone suggest what could be the problem with model or refinement? I will appreciate the suggestions. Thanks in advance John