There are a number of methods that might work. Let me suggest two. One would be to do a protein crash, i.e. precipitate the proteins in the sample under fairly harsh conditions. This will often release bound molecules, but not always. This is something that would need to be determined in validation experiments. You would follow the protein crash by LC-MS/MS.

A second way is to digest the sample with a protease, such as trypsin. This will usually cause a release of molecules bound to proteins. You would follow this step by LC-MS/MS.

Alternatively, you may be able to correct for the recovery by including a stable-isotope labelled internal standard.