It is generally understood that simulation predictions of building energy use, will by their predictive nature deviate slightly from actual real-life measured figures. Simulation and actual building in-use results may differ because of the assumptions made in the model, the quality of the data used, the quality of the simulation tool used, and how the building was commissioned, as well as many other factors. The value of simulation is in allowing whole building performance comparisons to be made between possible alternative design solutions that will guide the design to the optimal configuration, and in proving performance targets will be met before a brick is even laid. The arena of feeding back actual building energy performance data into a simulation model in order to calibrate it to what is happening in real-life is an area in which IES is particularly active on the R&D side.
See http://www.iesve.com/software/cloud-solutions/ergon for more info.