The program KiNG (which can be launched from the validation GUI, and is also available via the MolProbity server) will display the CB outliers graphically - represented by a sphere, the radius of which indicates the displacement from ideal position.
Since this is obviously a very high resolution structure, the first thing to check is whether these residues are in alternate conformations. It is quite common (although almost always wrong) for disordered residues to be split starting at CB, i.e. assuming that the mainchain is rigid and only the sidechain atoms differ between conformations. This tends to lead to CB outliers, because the mainchain actually needs to be flexible as well. It is much better to simply split the entire residue, regardless of how much you think the mainchain needs to move; by the time you can see clear alternate conformations you usually have so much data that the extra parameters won't be a problem. (It's often a good idea to split the adjacent residues as well, since their backbones may also need to flex a little bit to accommodate the central residue.)