Instead of testing newly proposed insulin dosing algorithms directly in clinical trials, it is nowadays a common approach to first perform simulation studies using complex physiological models of the human glucose metabolism (so-called in silico evaluations). The most prominent example for this type of models is the FDA approved UVA/Padova simulator. As an alternative, several authors have developed methods in recent years that try to extrapolate the effect of a modified insulin therapy using real measurement data together with simple, linear models of insulin action. These methods will be referred to in the following as "Deviation Analysis" approaches. Even though Deviation Analyses and in silico evaluations with physiological models seem to be very different approaches at first glance, it is shown in this paper that this is not necessarily true. Deviation Analyses are not restricted to using linear time-invariant models for describing insulin action. On the contrary, it is even expected that physiologically more accurate models of insulin action increase the reliability of Deviation Analysis results. It is demonstrated in this paper how even a very complex model like the UVA/Padova simulator can be used as a tool for performing Deviation Analyses. The corresponding methodology represents a step towards bridging the gap between classical in silico evaluations and Deviation Analyses.