OTOH ... he actually has pictures, and his pictures show what-I-would-call over-saturation on the later models, but that's just a personal preference.
OTOH ... jamma-nation-x goes by board revision which is significantly more helpful. At best, AES serials give you a vague indication of the board inside. At worst, it doesn't help at all. SNK used to give refurbished boards new serial stickers. It's not uncommon to find AES consoles with serials that don't fit with the board revision.
That guide makes it look like they *all* need modifying, with either trace cuts or recapping, and it seems to imply that they're *all* A-rated after the different fixes.
Do you have any info that shows that rev 3-6 is better than any of the others after the "fix" on that site?
The problem is almost every AES revision was poorly designed in some regard, like the reversed stereo audio on the AES3-4 or the vertical lines in the AES3-6.
The AES3-6 has very clear RGB with correct colors, perfect saturation. The problem is the vertical lines, but it's an easy fix. The AES3-5 is good too if you can find one without the weird checkerboard effect in the video. The AES3-6 is also very common, unlike the older board revisions, just find a high serial AES (150k+).
The problem with the earlier boards is the color saturation is either too weak (AES3-4, AES3-3) or too strong (NEO-AES with or w/o daughterboard).