Simulations are highly subject to the old GIGO limitations. Further, the quality of simulation software to date has not been up to accurately predicting complex systems with communications and distributed control schemes involved. The historical approach has been "divide and conquer", estimating the impact of each sub-part of a larger system. Still, such approaches have yielded only several-order-of-magnitude accuracy because of the significant effects of simple arrangement and packaging issues. Until we can get past the "We must be sure it is right the first time, but you cannot develop a circuit or packaging using costly features!" mentality, it is my considered opinion that simulation for EMI/EMC will only expose the efforts of designers that are not meeting the "motherhood and apple pie" design parameters.
EMI simulations will help a lot in reducing the R&D cost and time. The simulation tools for the automotiive industry are in the baby phase and lot more inputs are required from practical live field studies. These are on mobile and their ambience varies in a wide range it is more complicated to predict its nature using simulation alone.