The importance of simulation assumptions when evaluating detectability in population models