Building the model introduces inaccuracies into the estimating process. This is inevitable, and also beneficial. You are trading off model simplicity for accuracy. Doubling the effort on the model may give you only a slight increase in accuracy. Your experience will tell you when to stop refining.
