One possibility is low affinity: IgM is decavalent. If the antigen is coated on a surface at sufficient density, multivalent binding of the IgM profits from avidity effects: it can only dissociate if all binding sites dissociate. On the other hand, if you coat the antibody and try to capture a monovalent antigen, you have no avidity effect and therefore the much higher off-rate of a single binding site. The antigen may wash off too quickly to be detected.

With your sandwich assay what are you using as the detector reagent? same IgM monoclonal antibody (HRP or biotin labelled) or a polyclonal antibody? If it is the same IgM then you will still have the same issue as mentioned by Annemarie. Also, when coating with the full length protein you might be denaturing the protein (as a result of binding to the solid surface and exposing the peptide antigen determinant to the antibody while when the protein is in its native form in solution the peptide antigen may not be accessible to antibody. This is a common problem when using antibodies to peptides and use them to detect the native protein. It may work nicely doing SDS-PAGE western where the protein is denatured but not when the protein is in its native form

For the reason Annemarie gave you my have more luck with IgGs. Did you have any other positives that might be IgGs? As IgMs are so difficult to purify and generally of too low affinity to work as capture agents, I would screen the hybridoma cultures instead (or additionally) with IgG-selective secondaries. Only if I got no IgG positives would I work with an IgM.