Honestly, I think the best way to define simulation is to compare how you would resolve an event in all three of the main approaches - gamist, sim and narrativist.
So, let's have a thought experiment. A character finding and removing a trap. Pretty straightforward, bog standard action.
1. Gamist approach:
Player declares he's looking for traps, rolls dice to succeed or fail. If he succeeds, he finds the trap, then rolls dice to remove it. The nature of the trap, the details of the trap, what he's doing to find the trap or remove it don't really matter. They can be completely ignored. Traps are essentially like you find in something like the old Baldur's Gate game where you click the find traps button and a red outline appears where the trap is, then you try to remove it. There's no narrative here, particularly. You certainly can narrate the event, but, again, it doesn't really change anything.
2. Narrative approach:
Player declares he's searching. That indicates that the player is now interested in finding something here. A success might mean he finds something without any complications. A failure might find something or might result in the DM adding something to find that just blew up in the character's face. A lot of this is going to depend on the system, table, players and GM. Of the three, I think that nar games create the most idiosyncratic games because there are just so many different inputs to cover.
3. Sim approach:
Player declares he's looking for traps. Now, in earlier D&D, before you had thieves abilities, the player would actually have to declare what he was doing, how he was looking, where he was looking, etc. So, there's a strong element of sim here. There's a pretty direct correlation between the player's declarations and the narrative of the game. Later RPG's tended to gamify this a bit more because the whole Mother May I nature of declarations was typically deemed as too much of a PITA. So, again, you might get a system where the hidden trap has Hiding HP (I'm making this up as I go, please be kind

) and the character has Searching HP. A contest ensues as to whether the player can find the trap and disarm it. If the trap wins, the trap is set off. Again, there's a pretty clear line of progression here where the actions by the player can be directly narrated and won't be countered by future events. The whole Process Sim approach. Note, there are other approaches here that might work as well.
In any case, the three approaches (and of course you can certainly hybridize these as well) have advantages and disadvantages. The Gamist approach is fast. You get results very quickly. At the cost of not really generating any sort of narrative other than whatever the group decides to plaster on after the fact. The Sim approach has the advantage of generating a narrative. We know (within certain limits) what's going on. We can make narrations that are tied directly to the game and know that these narrations very likely can't be retroactively rewritten. The Nar approach has the advantage of being very creative and engaging. It has the disadvantage of being very creative and engaging. Sometimes I just want to roll dice and get on with it. Futzing about watching Dave and the GM waffle back and forth about some quantum trap can be fun, but, can also be well, a whole barrel of not-fun.
But, when all's said and done, none of the different approaches are superior. They're just tools.