Let me preface this post by saying that I'd be fine with either bans OR DSR(M), just not both. I'd personally prefer DSR(M), but I'd be fine with bans if DSR(M) was removed.
The goal of a set should always be to play the 3 (or 5) most fair and evenly matched games possible.
I disagree. What it sounds like you are saying is that we should attempt to balance each individual game rather than the set as a whole. That sounds good
in theory, but in practice I just don't think it can be done. I think the goal of a set should be to be balanced as a whole. Stage-based imbalances will always exist, but if we know in which direction the imbalances point, we can still attempt to balance the set as a whole, and probably better than we can balance each individual game. Let me outline my issues with your theory:
ISSUE #1: Suppose we are trying to balance a certain MU. Suppose this MU is extremely double-edged, like the Marth-Fox MU (for the sake of argument, humor my ignorance of this MU), where every stage gives either Fox or Marth some sort of advantage, however small. If we try to balance the set by balancing each individual game, we'd end up picking the stage that confers the smallest advantage (let's say DL gives Fox a small advantage, smaller than the advantages conferred on either character by all the other stages--again, just humor me) and forcing every game in the set to be played on that stage. The advantage given in each individual game is small, but the net advantage over the course of the set is decidedly in Fox's favor. Even if DL takes a Marth 50-50 Fox MU and makes it a Marth 45-55 Fox MU, the Marth could only expect to win a bo3 42.5% of the time against a Fox of equal skill when he ought to win the set 50% of the time. Sure, Marth lost on the "most fair" stage twice, but the "most fair" stage still wasn't all that fair.
ISSUE #2: Suppose I play Falco and am playing a set against a Fox. Suppose we're pretty evenly matched, I win game 1 by a hair and he wins game 2 by a hair. I take him to FD for game 3 because it's an even stage (just humor me here, I don't play Fox or Falco) and I think he'll stick, and when he does I pull out my pocket Marth and CG him to death four times in a row. He'd be completely defenseless against that because he's still banning stages under the assumption that he's about to play a Fox-Falco MU, which is the only assumption he can reasonably make unless I have a well-known and chronic history of CPing Foxes to FD and then pulling out my pocket Marth when they stick. In other words, it is impossible to balance the stages around the MUs as long as we keep picking the stage before we pick our characters.
ISSUE #3: Suppose we figure out a way to overcome Issue #2. This would pretty much have to involve picking characters and then having some impartial third party pick the stage. Assuming that said third party has no biases or non-mainstream opinions on MUs, what parameters could this third party use to pick the "most fair" stage (or stages, if we're following my philosophy)?
Can the third party use current metagame and/or "generally accepted" opinions on the given MU? Even if there were unanimous agreement on MU percentages and such, that's problematic because it precludes metagame evolution and also because it only applies to the highest level of play. If a Marth isn't good at CGing, FD's lack of platforms might not help him so much vs. an equally-skilled Fox who could possibly exploit FD's large size. If a Fox isn't good at shield-pressuring, PS's low ceiling might not help him so much vs. an equally-skilled Puff who could possibly exploit PS's low platforms and lack of a center platform. The fact is, the vast players are not at the peak of the metagame. There are only 10 players in the top 10, after all.
Can the third party use player skill? If he could do that, there would be no point in playing out the set. It'd only be somewhat feasible if both players had exactly "equal skill," which, despite the fact that I use it all too frequently, is an extremely slippery term which practically requires the two players to be literally identical. Let's suppose that there are two equal players that aren't identical. For example, Falcon A specializes in comboing and Falcon B specializes in grabbing and techchasing, and both are identically mediocre in both each other's specialties and in every other aspect of play. Since small stages and low platforms favor Falcon A whereas large stages and absence of low platforms favor Falcon B, this is effectively Issue #1. We might propose PS (large stage, low plats) and BF (small stage, no low plats) as theoretical compromises, but even these stages are going to be imbalanced. I would think that most people would attempt to compromise by going BF, then PS, (then BF, then PS for bo5) but eventually the third party would have to make a decision as to which stage to go to for game 3 (or game 5). A responsible third party would, in that case, decide the stage--and therefore the expected outcome of the set--with a coin flip, an action that essentially could have saved a ton of time if he'd just done it before game 1 and decided the set right there. After all, the point of a set is to be balanced to the point where two players of equal skill have only a 50-50 chance of winning the set, and the best way to do that is a coin toss.
Basically, it comes down to what Kal has been saying all along: There is no "fair" Melee. There is no "balanced" Melee. There is no "perfect" Melee. It doesn't exist and we can't create it (unless God decides to personally descend from Heaven and tell us how we should program Project: Melee). All we can do is get rid of blatantly, disgustingly unfair situations as best we can, and we've pretty much done that by narrowing the legal stagelist down to 6. Our SJR (skill-to-jank ratio), at least with regards to what we currently know of the metagame, is as good as it's ever going to get. If we narrow down the stages further, we remove more skill than jank; if we reduce these six down to just four, we remove from relevancy the sorts of skills that flourish on the other two. And I don't think that's something that anybody wants.