In general, I agree with you. You flexibilities coming out of NozzlePro or another FEA should be based on the junction itself.
As for the nozzle projection itself, I would model that as pipe. Really, once you are outside of the nozzle shell interface we as pipe stress really don't care where the purchasing folks put the flange connection that separates the "pipe" from the "vessel". It all behaves like pipe to us. Though, to be fair to your coworker, that stub of pipe between the shell and flange is probably not going to make a big contribution to system flexibility compared to the rest of the system.
As for the vessel itself, we tend to model a rigid element from shell point to the centerline of the vessel. I then use pipe elements to model the vessel for graphical clarity, but I will use an absurdly large wall thickness to minimize it's contribution to the system flexibility. (if the drawing shows a 5/8" shell, I might code it as 5" thick) My reason for this is not a concern for "double dipping" per se as to avoid dealing with the uncertainty - internal trays, tube sheets, etc are going to affect the stiffness is ways that are too difficult to adequately predict for consideration in Caesar, so I code them out of the design altogether.
Once I've considered local flexibility of the nozzle, if I still can't get the system to work, some unknown extra bit from the rest of the vessel is not likely to negate the need to flex up the overall layout some more.
_________________________
Edward L. Klein
Pipe Stress Engineer
All the world is a Spring