Yeah, that's exactly what I was saying. The only absolute rule for SQ mixing was that you couldn't have identical information in both rear channels (ie center back position) because it would cause the SQ encoder to create pure vertical modulation, which would make record needles skip out of the groove. Beyond that the other 'rules' are more like "strong suggestions" for optmising decoding, things like not having reverbs from instruments in the front appearing in the back (and vice versa) and not having stereo images along the side walls (ie in FL & RL or FR and RR).
It's my understanding that the CBS studios that were quad equipped (primarily speaking about NY here, but probably their studio in SF as well) had a monitoring system that allowed the engineer to listen to what the mix would sound like decoded in SQ at the flick of a switch. So they'd basically be listening to a chain of 4 channel output -> SQ encoder -> SQ decoder -> 4 channel playback. This allowed them to see how their mixing decisions fared during the encoding/decoding process, and change things that didn't work. Also CBS had a vested interest in promoting SQ as the best quad system (since they were one of the patent holders) so they made mixes that maximised the separation between channels, especially in the early days ('72/'73) when SQ decoders were pretty much hot garbage. Think of stuff like Janis Joplin Pearl, Sly & The Family Stone Greatest Hits etc. - what Adam would call 'four corner' mixes, because that kind of mixing style got the most out of the limitations of the SQ process. So with a lot of the Columbia/Epic quads, I think the SQ LP was their main consideration, doing a mix that would get the most out of the LP format, and the Q8 was really an afterthought, basically just a discrete 4 channel capture of the work that was done to create the SQ-optimised mix.
A&M on the other hand, just did their quad mixes as they pleased (as that article I linked to says) and then just SQ-encoded them when they were done, so it's no surprise that some of the quad LPs don't decode as well. It's not because the mixes were bad (the Wakeman and Quincy Jones quad mixes are some of my favourite quad mixes ever) it's because they didn't let the limitations of SQ influence how they did their mixing, as Soundfield's post above explains better than I ever could. As I said before, I think this is part of the reason they later moved to CD-4 - with that format you didn't have to worry about what effect your mixing decisions would have on the decoding process, and I think that suited their style and philosophy. It's a shame A&M shut down their quad program in 1975 - much like RCA and WEA, they were just hitting their stride in terms of fully-immersive, detailed and nuanced quad mixing just as they pulled the plug on it.
One of the things I love about quad is how many different approaches there were to mixing, and how the talented engineers that did it often took the limitations they were working under and used them as inspiration. Without the "rules" and limitations of SQ we may never have had some of the crazily discrete super dry CBS quad mixes that many of us love so much. Imagine if the majority of quad mixes were 'big stereo' with the same vocals and instruments in all 4 speakers - I don't think we'd still be obsessing over them 40+ years later, instead they'd probably be consigned to history's dustbin along with things like Capitol's "duophonic" fake stereo of the 60s.
