Proteins undergo post-translational modulation (glycosylation, acetylation, binding with other proteins) as well as proteasome-dependent degradation, both of which cause several different MW (molecular weight) proteins. Further,  the diversity of mRNA due to alternative spicing results in the different MW products even before the translation.

[Typical Examples]

The smear-typed high MW suggests the glycosylation.

The ladder-typed low MW suggests ubiquitin-proteasome dependent degradation. 

The sharp several different MW bands should be checked even in terms of alternative slicing, binding with other proteins or formation of dimer, or just the non-specific binding.

You have to keep in mind that siRNA-mediated knockdown can distinguish whether the band is specific or not.

The bands related to different molecular masses obtained with different antibodies may result from the specificity of the antibodies.

Generally, if you used polyclonal antibody raised against a whole length protein, you should obtain bands corresponding to all form of the processed protein. More confusing are polyclonal antibodied raised against peptides - in that case you visualise isoforms of the protein which contain the peptide used as an antigen.

Monoclonal antibodies may be accurately characterized. It means that the supplier should provide information about the isoform of the protein which is recognized by the reagent. Mabs may bind specifically to a protein with particular glycosylation pattern, to particular isoform or protein which was processed in a desired way. 

Do not forget about another possibility. Some vendors rely completely on customers to validate their antibody. We used to have a polyclonal rabbit antibody identify one 90Kd band where as the protein molecular weight is 130 Kd. We crop the band and found out it is not the protein we are looking for. Also, make sure the protein you are looking into share a similar epitope for the antibody.