Thanks guys!

Using a PLD to do memory mapped I/O or a uC with more I/Os available is certainly an engineering solution, but I am afraid that I may not have the luxury in terms of the product cost.

As to what Joe said, in my first failed attempt, I remember that I tried writing 0x00 or 0xFF to P0 after reading from the tri-stated buffer (74HC244) or writing to the latch (74HC573) to free up the internal bus and it still didn't work.