Hi John,
Just curious have you looked at the IS66WV1M16DBLL-55BLI at mouser? The 1M x 16 and a whole lot cheaper than the chips you are using and you would still only need 8 of them and they are TSOP 44 as well. You could map even/odd bytes with the LB/UB signals. I would assume most processors that are going to use the 16M are going to be 16bit processors so 16 bit transfers would be the norm.
BTW - if you want some help with the boards I would go in with you for a couple of the prototypes to do some testing as well.
Regards
Dave