I agree with the prior discussions as far at the actual locations of these pressure thrust loads, but suggest that the engineer must also keep in mind the intent of what he/she is attempting to simulate. Is it actual pipe stresses or resulting loads on equipment?

In the case of pump nozzle loads, the limit on nozzle loads in API 610 is indented to maintain the machines alignment, and is not based on pump casing stress. This limits any misalignment between the pump casing and the driver or gearbox.

If you then consider the pump casing as a totally rigid body it does not matter where the pressure thrust force is applied, either at the unrestrained expansion joint or split between the expansion joint and rear of the pump housing. Both will result in the same effect on machine misalignment.

When considering nozzle loads on rotating I would always recommend:
1) With an unrestrained bellows in the system, simply applying the pressure thrust force (based on the effective area the expansion joint x pressure) at the expansion joint connections.
2) Without an unrestrained bellows in the system, the pressure thrust forces are contained within the pressure containing metal (pipe and pump casing) and will not be seen externally, therefore no external loads need be included in the model.

I hope this as helped on the understanding of this issue