From a short introduction to immunohistochemistry controls:
Isotype Control
This control can be utilized when working with monoclonal primary antibodies. The sample is incubated with antibody diluent, supplemented with a non-immune immunoglobulin of the same isotype (for example, IgG1, IgG2A, IgG2B, IgM) and concentration as the primary monoclonal antibody. The sample is then incubated with the secondary antibody and detection reagents. These steps will help ensure that what appears to be specific staining was not caused by non-specific interactions of immunoglobulin molecules with the sample. Background staining should be negligible and not resemble specific staining.
I googled and found out that some pathological agents can bind immunoglobulins in a non-immune fashion, i.e. by binding to their Fc regions. But that still leaves me in the dark on "non-immune IGs".
Maybe it is an IG with which the secondary antibody will not bind?
Or maybe it's an IG that has no activity against the target antigen?