i know tis is a gazillion years ago, but i got a request for info on what i did back then, so im posting the Email reply i sent him. hope it helps other ppl.
Im amased this issue still ,generates interest more than 10 years after
First of you are never gonna get the FTDI232RL to work better than what i did back in 2006. That old generation
USB chips dident support bitbanging, and used many usb packet to write one bit. Thats why i took me 30 min to write an atmega30L.
back then i decided to use stkv500 protocol and build myself a small programmer with an atmega8 inbetween as a data buffer (to circumvent the problem).
but times has changed, the new generation ftdi chips called FT232HL, supports true bitbanging, and i think if you change my original design to this chip. It might just make full speed ponyprog programming possible. It just never had time to try this myself since, im using my stkv500 programmer.
i have uploaded the diagram for you, you can get it here https://postimg.org/image/3nv3dn3gb/
Hope it helps you.
Freindly greetings from Denmark