Indicators and "boolean processing"

I am trying to assign the result of a condition to an indicator.  I have
done this in the past, like the following:

     d CorrectRec      s               n   inz
     d ContainsEM      s               n   inz
     d EMisFirst       s               n   inz

      /free
       CorrectRec = (ContainsEM And Not EMisFirst);

This works with no problem, and the indicator CorrectRec is turned on only
if ContainsEM is on and EMisFirst isn't.  The same thing can be accomplished
by more code:

       If ContainsEM And Not EMisFirst;
          CorrectRec = *on;
       EndIf;

I am wondering, is there a level of complexity you can't cross using this
technique?  I find nothing in the IBM documentation regarding a limit as to
nested logic, yet when I extended the program to look at more than one
condition, apparently I got too complex.  When I use the following formula:

             CorrectRec = Not ((ContainsEM And Not EMisFirst) Or
                          StudentCount <> ModelChgCount);

and when I attempt compilation, at the end of the compile listing, I get the
following error:
                 C o d e   G e n e r a t i o n   M e s s a g e
s
An error occurred during
translation.
 * * * * *   E N D   O F   C O D E   G E N E R A T I O N   M E S S A G E
S    * * * * *

And I get a message in the QSYSOPR queue:
* Software problem data for QRN4CPP has been logged. Refer to help text for
*   additional information.
* Software problem data for QBNCRTMD has been logged. Refer to help text for
*   additional information.

Has anybody run into this?

--
"Enter any 11-digit prime number to continue..."
"In Hebrew SQL, how do you use right() and left()?..." - Random Thought
"If all you have is a hammer, all your problems begin to look like nails"