MATPG corrupts data at offset X'B0' ?

I am observing a problem with MATPG.  For one OPM program, if I compare the
MI creation template I get from MATPG with the creation template that I can
print from service tool display/alter/dump, I see the MATPG instruction
corrupts the first bit at offset X'B0' in the template, setting it to '1'.

(If you don't want to bother to write a program to run MATPG to test this,
you can just look at the output of a DMPOBJ (which I think runs MATPG) on a
program, and compare that to the service tools display/alter/dump of the
OPM creation template, and you'll notice the different bit value at offset
X'B0'.)

Now for our programs created through QPRCRTPG or QPRROOTP, this problem
seems to not bother too much, since QPRROOTP begins the instruction stream
component at offset X'100' and the ODT component after that, so we really
don't care about offset X'B0'.

However, today I am trying to run MATPG on old programs created from a
non-QPRROOTP translator (I'm told it's "PLMI", whatever that is), which
placed the ODT component at offset X'A0' and the instruction stream
component after that.  Guess what, in all of these programs, the fourth ODT
entry is always returned with the first bit corrupted to '1' (since the
fourth 4-byte ODT entry starts in offset X'B0').

Yesterday I thought the problem was some new 'type-E' constant thingy
instead of a 'type-6' constant thingy, but, today I realize the problem is
occuring in all programs where the ODT starts prior to offset X'B0'.

If I try CHGPGM FRCCRT(*YES) on these programs, the command fails with
MCH4202 "ODT syntax of entry 3 not valid."   However the QSCMATPG &
QSCCRTPG  work okay on these programs.  What gives?