Instead of using the spectrum to describe the source, you can use the phase-space data after the beam has passed the jaws. This will also give information about the spatial distribution of the particles generated. At this point you can modulate the beam to suit your needs.
International Atomic Energy Agency (IAEA) has initiated an effort led by Roberto Capote to create a database of phase-space data of linear accelerators (and Co-60, and electron linac). Within this initiative the phase-space data of many linacs are provided, for instance Varian TrueBeam or the Elekta precise. See their website for more information: https://www-nds.iaea.org/phsp/phsp.htmlx .
Now you can use this phase space file and further modulate the beam to create the specific treatment plan you need. Since you are working with MCNP there might be a bit more difficult to import the phase-space data, but definitely ways to work it out exist; some methods described on the IAEA phase-space database website.
If you start your simulation with initial electron beam, you can choose monoenergetic point (or 1 mm diameter) electron source hitting the target. For the breast plan it should be very good approximation. The energy of the electron source should be find from PDD agreement for standart 10x10 field (measured and simulated) or just take the value from previous works (for your linac). Energy vs intensity distribution of electron source is not relevant for your kind of simulation and will not affect the results.
If you work at an institution using Varian equipment, you will have access to https://varian.force.com/ where Varian share detailed descriptions of the heads of the linacs as well as phase space files. Hope this helps, good luck!