The inability of otherwise successful dynamical models to reproduce the "HBT radii" extracted from two-particle correlations measured at the Relativistic Heavy Ion Collider (RHIC) is known as the "RHIC HBT Puzzle". Most comparisons between models and experiment exploit the fact that for Gaussian sources the HBT radii agree with certain combinations of the space-time widths of the source which can be directly computed from the emission function, without having to evaluate, at significant expense, the two-particle correlation function. We here study the validity of this approach for realistic emission function models some of which exhibit significant deviations from simple Gaussian behaviour. By Fourier transforming the emission function we compute the 2-particle correlation function and fit it with a Gaussian to partially mimic the procedure used for measured correlation functions. We describe a novel algorithm to perform this Gaussian fit analytically. We find that for realistic hydrodynamic models the HBT radii extracted from this procedure agree better with the data than the values previously extracted from the space-time widths of the emission function. Although serious discrepancies between the calculated and measured HBT radii remain, we show that a more "apples-to-apples" comparison of models with data can play an important role in any eventually successful theoretical description of RHIC HBT data.