Thanks for the response. It is an R-Tag v1.0 board with v1.1 firmware on the chip. (I thought that R-Jtag was a Coolrunner with a new program.) In previous posts, I stated that with only Xell Reloaded flashed to the nand, I got successful boots 5 out of 5 times. However, when I built the full image and wrote that to the nand, successful boots were few and very far between. So, with no changes other than what is written to the nand, it cannot be a soldering issue, where it boots perfectly in one, but fails in another. The same soldering exists for both. I will also re-iterate, that it boots much, much better with a falcon image, than with a zephyr (if 1 out of 50 attempts can be called better). As far as optimization goes, I have stated that I have tried every possible combination of options, even selecting sw 7-8 off-off, off-on (for zephyr) and on-off (for falcon. Just 1 option left, using the top-side rst point with my own soldered flat cable. Will try that soon, but I don't expect any significant difference.
Edit: Another question - With no jumpers on the voltage selection pads, what is the voltage on the center pad? I measure 1.9V, same as the so-called 1v8 pad. The pads are NOT bridged, as verified with an ohmmeter, which measures 480K between them.