Yes, it is indeed possible to program the GPIF II to accept only every alternate byte.
Thank you. I think I will probably try to generate the required state logic for the GPIF II before I commit to a PC layout, but I just wanted to be sure I wasn't wasting my time trying. This looks like a very capable and flexible part, but it's completely new to me. I'm used to doing a lot of glue logic in CPLD or FPGA and then using a more standardized interface for other chips.