So the big tl;dr with nSequence vs any other scheme is that only nSequence is based on per-txin-scriptSig information, or basically it's the only scheme(1) that's CoinJoin-compatible. The reason why it's compatible is that all the information about where the "color" of the txin is meant to go is in the txin, rather than any other part of the transaction, which allows the rest of the transaction to be specified by other individuals.
I'm not sure whether you mean that it is compatible with a particular existing protocol, or whether it is compatible with p2p coin mixing in general.
I believe that 1) we need to use color-aware p2p coin mixing for colored coins; 2) in that case it doesn't matter which coloring scheme you use.
First, I should note that I did research on p2p mixing
before it was cool about a year before gmaxwell described CoinJoin, so I don't know what terminology is currently in use... But anyway, I describe quality of mixing in terms of entropy, i.e. a measure of uncertainty. E.g. if attacker is able to trace a particular txout to a set of 1024 txouts, each of which can be a source of money equiprobably, then mixing entropy is 10 bits. Bigger is better.
Suppose there are colored coins in a CoinJoin transaction, for simplicity we can assume that there is one txin/txout with each color. Obviously, attacker can trivially connect txout to txin. So we get two things out of it:
1. there are no mixing entropy gains for colored coins
2. mixing entropy gains are reduced for non-colored coins
So it is definitely better when mixing is color-aware... And if mixing is color-aware, then order-based coloring and its variants work fine. Basically, you just do this mixing thing for each color in separation (it's OK to reorder txins/txouts withing one color), and then you just concatenate smaller transactions together (txins_1 + txins_2 + txins_3, txouts_1 + txouts_2 + txouts_3).
So now
1. it is possible to mix colored coins too
2. everybody can easily understand entropy gain and plan accordingly.
Perhaps I'm missing something... If your point is that it is possible to re-use existing protocol (if it is possible, I'm not sure about that: we need an ability to modify nSequence after list of outputs is known), then I guess it is good because it is saves work, but doing color-aware mixing protocol is definitely better in the long term.
being able to easily hide who paid for a given colored coin is useful privacy, and increases the anonymity set for everyone else.
You can get same results by mixing your coins before you buy that colored coin.
But everywhere I read about colored coins people don't really get specific about how it works. 
