I believe that what you are getting at is that there is a specific structure of P+/Nwell/Psub that is used for the "bipolar", so you are asking why use that structure rather than simply a P+/Nwell "diode". Here is my take on this:
1- The "bipolar" will simulate more accurately than the "diode", since it will include the substrate current that is probably not modeled for the "diode".
2- There usually is a specific structure for the "bipolar" that has characterization data available. When building a bandgap structure, the good characterization is needed in order to properly determine the tempco of the Base-emitter voltage.
3- The additional structure of the bipolar should help prevent current injection into other substrate tied devices.
There is, of course, nothing preventing the use of a P+/Nwell diode in your application.