I can't help a lot because I have never run AEGIS and rarely run the
cereal models, but here are some general comments.

Tomato and Soybean use the CROPGRO model, all of the cereal
models use CERES.  Although they use similar input files, the
underling code of the models are quite different.  You could be
looking at a difference between the models.

The M6101 error means that the model is trying to use an invalid
value in a calculation - usually where an variable has a value of "not
a  number" or infinity.   This might be from having a -99 where the
model expects either a 0 or an actual value.  More likely a number
has gotten too large or too small and the model doesn't recognize it
as a number any more.

Does this problem occur on the first year of the simulation or only
after several years of runs?

If the problem occurs on the first year of the run, I would look
through the "X", experimental file, for a field that is not in the correct
place/column (somewhere there is a field that is not the proper
width).  This would have to be unique to the lines for the treatment
numbers that had the water and N options turned off.

If the problem only occurs after several years, then  I would assume
that this is an "accumulation" problem where some value keeps
accumulating year after year and eventually grows too large (too
many digits) to be saved properly or read in for the following year.
You might be able to track this down by looking into the OUT files
generated for the last run - see if there is a column that has "***"
instead of numbers.

Have you tried running your simulations in DSSAT v4?  If you are
seeing the problem only after several years of simulation, this might
be your best/only solution.

Stu