You need to find out what the problem really is. It is clear (at least to me) that nobody has a clue.
Write the absolutely smallest program you can, one that does not in any way depend on the timing of instructions etc and use the absolute minimum of sfrs.

If you have a scope (since you seem to be producing, you should) make a program in assembler such as
Org 0
here: setb p1.0
clr p1.0
sjmp p1.o
end

and scope p1.0

Does that work? Whether it does will tell a lot

Erik