The power factor is only unity assuming the elements are purely resistive. They're probably close enough that it doesn't matter, but you'd have to consult the specifications to be sure (or test it).
As to why one phase has a unity power factor, that's really better answered as why the other two don't.
If there were 3 (or 6 or 9, etc) elements in the heater, each on a different phase, either P-N or P-P (all the same way, of course), then the overall power factor would be the same, and virtually unity (to the extent the elements are not reactive). In the case of P-P, the currents on each element are phase shifted from the voltages on each phase wire (as in any P-P connection), but when the currents meet at the junction, one leads and the other lags, and they cancel out (some current just flows between them) to no phase shift.
In the case of 4 elements (this water heater) wired P-P where A-B gets 2 of them, the above description of cancelling out only applies to phase C because it has the same number of elements wires in each direction (one to A and one to B). So it has the same current leading as lagging. The other 2 phase wires (A and B) have the angular imbalance because there is one element in one direction, and two elements (more current) in the other direction.
If you had 3 such unbalanced 4 element water heaters, you could stagger them 3 ways to get a net total lower load than if you just wired them up all the same, making it equivalent to a 12 element water heater system. If you did have them, doing this would be good at least to get the better power factor as seen beyond the point where their circuits meet.
You might be tempted to favor phase C for some other loads a bit more due to the lower loads from this water heater on that phase. But that wouldn't be a good idea, and your AHJ is probably enforcing that. Consider the water heater fails some day in the near or far future. You're not around, and the next guy to come along to do the wiring has no idea how you balanced other loads to match this unbalanced load (even if he does understand it is an unbalanced load). He might end up wiring the phases differently (he's got a 1 in 3 chance of getting supply phase C connected to the terminal with only 2 elements). If he gets it wrong, your balance with other loads is all thrown off.
So basically, for the purpose of calculating building loads, figure the water with each phase having the current equal to the worst cases. Only if you were putting 3 such heaters right next to each other, and leaving everything well documented, might you consider staggering the phases between them to get a slightly lower load. And even then the AHJ might not allow you to calculate everything that way (and I can see why not for the safety reasons).
FYI, I'm not an electrician. I design/manage computer/network data centers, and do have to deal with things somewhat like this surrounding UPSes and such, sometimes. Then I hand off the work to an EC who, if I did my end right, won't have to come back and tell me the code, or face a red tag. So I learn a lot of this stuff so I can get it right the first time and not have to re-design things all over.