You're singing to the choir, my friend. I don't think it does, either. I don't think it _can_, and for the most part, I don't think it _should_!
I agree with Duke and dsatow here. The quotes are for emphasis.
Anytime you are creating a simulation (or model if you prefer), you must make some simplifying assumptions. If you could recreate reality without making some simplifying assumptions, you wouldn't be creating a simulation/model--you'd be creating...well...reality.
The trick is to find out what you can simplify (or take out) and still create an accurate enough simulation.
It reminds me of the old physics joke where a farmer goes to his physicist friend and asks them if they can find a solution so that his cows give more milk. The physicist goes away for a couple of days and comes back saying that they have a solution, but it only works for spherical cows in a vacuum.
While the gist of the joke is that cows aren't spherical or live in a vacuum, the reality (pardon the pun) is that, if the shape of the cow and its respiration isn't important to how much milk it gives, then the model could be valid. (I realize a cow that can't breathe, like one in a vacuum, isn't likely to give milk, but if the breathing doesn't contribute to how much milk it produces (only keeping it alive), breathing can be removed from the model and being in a vacuum doesn't matter (in this case).)
So, it seems to me that the crux of the problem is what simplifying assumptions you can or need to make regarding darkness and light and whether or not the model it produces is good enough for what you're trying to model. It's not likely we will ever find the perfect solution--just one that's good enough for an individual GM or their table.