First is that "I make a melee attack" no more defines an attack explicitly to me than "Come and Get It" does.
This is why I think the longer combat rounds of TSR-D&D work better. Both WotC- and TSR-D&D use very abstract combat systems, so much so that the details of your "melee attack" are unimportant. When you combine that level of abstract resolution with 6-second combat rounds, I think it's
easier to ask what, exactly, is my guy doing?
I think that's why, when I play pre-4E D&D, I usually say "I attack the orc" without going into detail. Maybe there's some level of detail if I'm using Power Attack or whatever, but that's limited to what my PC is holding in his hands.
Because of that, one feature of play that I experienced was that the DM shouldered the responsibility of narrating what was happening. A simple, made-up example:
Player: I attack the owlbear!
(roll, hit, 4 damage)
DM: You stab your spear into its side and leave a thin line of blood!
Player: Wow, my guy is awesome.
If, as I begin to suspect, the demands of "association" and "immersion" are that the player should have moment-to-moment control of every significant action the character selects - in the sequence as they occur in game-time - then I don't really see a good way to make such a system both simple and in any sense believable to anyone who has a vaguely realistic model of "fighting" in their head. It might work - barely - for the old Hollywood model of "let's stand here and take it in turns to whack each others' shield" "combat", but not really for anything beyond that.
I'm not sure that's true, but since I'm not particularly demanding of associated mechanics, I could be wrong. Let me touch on the "lethal ballet" for a second and get back to this point.
When I designed the combat system for my 4E hack, I wanted players to make choices that leveraged the fiction to their advantage - or disadvantage, in the case of poor play. I also wanted to picture the scene in my head. (I think that's immersion but I am not sure.) This doesn't demand associated mechanics; it demands (certain detailed) fictional positioning having influence on resolution.
This is why I like 4E - I think that the system makes it easy for fictional positioning to influence mechanical resolution. Dissociated mechanics may play a part here, because the player creates, at run-time, the connection between the game world and the mechanics. Which means that you can create your own fictional positioning - fiction that's relevant to you, in the moment, and can make as much sense as you want it to - instead of relying on the mechanic's connection to the game world.
Anyway. I'm playing my hack with a guy who is really into the medieval martial art recreation scene. After our first session he noted that my system got the "beats" of combat right, for him at least. (I think he had an Italian term for it.) After a few sessions, though, he went back to saying "I attack the gnoll" instead of using his greater knowledge of combat to his advantage.
After playing with this guy for a while, I think he went back to "I attack the gnoll" instead of more detailed actions because he didn't want the responsibility to narrate what was going on. He wanted to make a simple choice - "I attack the gnoll" - and then have the DM describe his character's actions.
I think this creates a reward system for a certain type of "Right to Dream" play. You make a character, but you don't want to have to risk that character's integrity by possibly screwing up with your choices - you want your choices to be low-stress. You pass the ball to the DM, and a good DM will "get it right" and describe your character in the "right" way - passing the ball back to you. Then you express your character well enough for the DM to pick up on what your PC's about, and round and round it goes.
That's what I see as the main benefit to associated mechanics: you aren't forced to create compelling fiction moment-to-moment during play. The mechanics do that for you.
I think that's a pretty controversial statement. Since I don't naturally "get" associated/dissociated mechanics, I think I could easily be wrong, especially if someone who naturally gets the theory says so.