Results from a detailed investigation into the effect of modeling assumptions used with the RSI method to compute broadband interaction noise downstream of a turbofan engine's fan stage are presented. The modeling assumptions that are considered include the use of a Green's function to obtain the exhaust noise from the unsteady vane surface pressure, the implementation of a 2D vs. 3D vane model, and the form of the turbulence velocity correlation function. Calculation of the duct acoustics via the Green's function is shown to be robust when one selects the frequencies used for the calculation such that they do not coincide with a duct cut-on/cut-off edge frequency. The unsteady vane response calculated by strip theory is found to be different than that predicted with a three-dimensional vane model. However, it is not clear yet how these differences specifically impact the predicted exhaust noise. Inclusion of the inhomogeneity of the turbulence across the passage is not so important because the average passage value provides good results. The form of the correlation function used to model the inlfow turbulence is shown to have a strong impact on the overall sound power level. Within the RSI framework, it is shown that using a common 3D spectrum (e. g. Liepmann and Gaussian spectra) but disregarding the k3 contribution gives results 20 dB lower than when the nontraditional RSI spectrum is used. The inclusion of the k3 effect with the common 3D spectrum within RSI leads to a difference of 10 dB instead of 20 dB; however, the physical argument for including k3 effects on each 2D vane strip is unknown.