Disparity in Results between FLUKA & MCNPX

Dear FLUKA experts,

I've been trying to cross-check some FLUKA calculations I have done
using MCNPX but I am having problems getting good agreement for some
scoring regions.

I have a beam dump with a carbon centre, surrounded in iron and then in
concrete. I have a spherical scoring region around this beam dump that
has been segmented using cones so that dose & flux can be scored in
different directions. I am finding that the neutron doses I calculate in
FLUKA for the directions opposed to the beam direction are up to ~factor
3 lower than calculated using MCNPX. Neutron doses in other directions
(through the beam dump and emerging from the sides) agree fairly well
and photon doses agree well throughout, it is only the neutrons where
the disparity is apparent.

I have also attached a simplified version of the same model, without
biasing and with only the carbon core. Here, all the neutron doses in
MXNPX are up to a factor of 2 higher than in FLUKA, whereas the photon
doses agree very well.

FLUKA calculations have been run using Fluka 2011. MCNPX using MCNPX
Version 2.5.0. All calculations have been run sufficiently long to
obtain SD < 5% in all scoring regions. I have also found that the
differences are less apparent in some other materials, i.e. iron and
aluminium. Note that flux-dose conversion factors in MCNPX are in
microSv/h and not in pSv so this has to be considered in normalisation
to compare with the FLUKA.

Any input would be most helpful, many thanks!
Lewis.