Well, I've figured out what the problem is and I've got a solution.
This probably only affects those of you programming the OTP PIC12C508's
using the Parallax PIC16CXX programmer.
        - SPASM wants to write the reset vector address to 1FFF. This
          location is really the OSCCAL instruction (MOVLW XX where XX is
          the correction number)

        - for the JW part, after erase, location 1FFF holds FFF.  SPEP has
          no problem programming this location with the reset vector instruction
          so the JW part works fine.

        - for the OTP part, location 1FFF contains whatever the correction value
          is needed for the OSCCAL.  SPEP cannot program this location with the
          reset vector instruction  BUT IT DOESN'T COMPLAIN EITHER!!

        - So, when the OTP part powers up, the first instruction run is the
MOVLW XX
          in location 1FFF.  The next location run is memory location 0000,
which is
          the start of my subroutine space.  The subroutine is run and then
the RET
          instruction is encountered but where does it return to?  This
subroutine was
          never called so the chip never runs!!

        - By adding to my code at memory location 0000 the instruction GOTO
START, the
          OTP part now works.

        So there appears to be a bug in Parallax's SPASM.  It is writing the
reset vector
to 1FFF instead of 0000, It doesn't complain if it cannot program this
location (OTP have
this location already programmed), and when assembling the code, it doesn't
complain that
you are beginning your code at memory location 0000 which should hold the
reset vector.

        So, for those of you running SPASM 4.7 from Parallax, be sure to put

        org     0
        goto    start

        in your code or the OTP parts won't work right.

I'm glad I didn't keep trying to program OTP parts and waste a bunch.

Dave