To me, saying that model or a theory is wrong is like saying a hammer is wrong. A tool cannot be right/wrong or true/false. However it can be unsuitable for a task. In this case, the task is understanding reality.
It is the second part of the quote (follow the link above): "all models are wrong. Some are useful" -- you are talking about the suitability/usefulness of a model.
Exactly, if your model has all the complexity of reality... then you're going to have a hard time understanding it or answering any question unless it is historical rather than prediction. They key is in the model being useful locally, and a big part of that is knowing where it is valid (e.g. Newtonian Mechanics is great for most things until you exceed micro/Tera scales). Then you can choose the right model to solve the right problem or answer the right question. Trying to figure out what every molecule in a gas is going to do will drive you crazy, but measuring pressure/temperature/entropy will often lead you to insight.
In my mind this is one of the problems that moral philosophy has fallen into... rule utilitarianism works because we are about as good at predicting how others would feel as we are for ourselves (i.e. mediocre, but not terrible). Most modern moral theory is so complex in order to deal with various oddities that we're terrible at using it in even the simplest cases.
