I agree with peterb. With the balanced curents that you're seeing it doesn't seem possible that the individual phase power factors are varying so widely, and it seems very likely that some instrumentation hookup problem is causing your instrument to sense an incorrect phase angle between voltage and current, due to an unintentional phase shift in one or both of these signals.
As I mentioned above, it's important to check for proper ground reference to your instrument (there may be one ground input or three ground inputs) and verify no delta-wye connection on the plant potential transformers, if used. If you happen to have an ungrounded system, then you would at least need access to a neutral.
As peterb mentioned, there is a polarity of a clamp-on probe. There are two ways to attach it, they give current 180 degrees apart. This in itself doesn't seem to explain your results... a 180-degree phase shift would give the correct magnitude of power factor but the wrong sign (i.e. negative power factor... which the instrument probably automatically recognizes and corrects, since negative power factor in a mathematical sense would represent real power flowing out of the motor). But it can't hurt to check this.
In my first response I assumed that you don't have current transformers on your 440 volt system. But if you have them and are using them to sense current, then recognize that the connections of current transformers (delta or wye) can also cause a phase shift.
