I'm not an MCNP user, but I've done some Monte Carlo simulations.
Some common things to consider/check:

• If you're simulating from the point of an electron source incident on your target, remember the bremsstrahlung process is not that efficient. How many histories are you running?
• Are you implementing any variance reduction techniques? If say one in every hundred incident electron produces a photon, depending on your geometry there's finite probability that said electron will be directed toward your phantom setup.
• What does your scoring volume consist of? Have you defined a water phantom at your isocenter? Or are you attempting to score dose in air?
• How large is your scoring volume?
• Does your output give you any error codes?