What I would recommend is to increase the delay, especially if you have a high speed crystal.

Also, I would suggest that the data is loaded BEFORE the LCD enable sequence commences. I discovered myself that when enable rises from low to high, data is written.

Since you have the first half of the program working, the only piece of hardware you should be changing if any is your external code ROM (if any) or your crystal. Dropping your crystal speed can help.

Also, try programming the code on a new ROM.