Difficult geometry problem

I have been struggling off an on for the past few months to produce the
attached .inp file, from CAD drawings and specific information, for a
beamline configuration. Something is wrong with the geometry, and I am
at a loss to find it.

We have a neutral beam along X=Y=0, passing in and out of vacuum regions
through a charged-particle beamline. There are awkward angles (8 deg.
and a later 13-deg. bend). Some cylinder bodies are longer than they
really are, and then cut by infinite planes to fit as a way of trying to
avoid precision errors.

The file passes all of the Fluka geometry tests I have thrown against
it, including some with tiny steps, giving no errors. It can be run
through Flair without errors to generate a plot (a potion of the
configuration is attached). I can run 100 and even 1000 RAY particles
with no errors. However, Fluka crashes after just a few events when
neutrons are used, with extensive pages of MA ARRAYs and FPD ARRAYs.

I tried to see what I could get from the RAY particles. But the file is
not readable by the code given in the manual. With some extra output I
added, I get:
  File name?
ni30_n001_fort.10
 nrayrn,mreg,mlattc,mmat,ekin= 2 0 26 1054835810 2.80259693E-44
  Incomplete data on file about ray starting point

For the neutron case, the .err and .log files have nothing beyond the
usual standard information. The .out file is far too big to be useful
and can be easily generated, and so is not attached.

One wish: It would be really, really helpful if there was some
intermediate structure in the geometry, i.e., a complex region that
could be built and then added as a unit into the geometry stream with a
position and rotation (with all precision issues taken care of).
Perhaps some user subroutine link. For the 'AirReg2' region, I wrote a
separate program to handle all of the displacements and rotations, and
then entered the relevant data into the .inp file manually. Very
tedious, and surely prone to many errors! In the end, I was thrilled to
find that the X=0 line was exactly where it should be. Yet, there must
be a better way; e.g., something that one might be able to apply XML
coding to.

I'm not looking for someone to spend a lot of time in the ,inp file, but
I need some good suggestions. Maybe the solution will be very simple,
but I have not found it.

Thank you,
Joe Comfort