In a recent experimental study the in-cylinder spatial distribution of mixture equivalence ratio was quantified under non-combusting conditions by planar laser-induced fluorescence (PLIF) of a fuel tracer (toluene). The measurements were made in a single-cylinder, direct-injection, light-duty diesel engine at conditions matched to an early-injection low-temperature combustion mode. A fuel amount corresponding to a low load (3.0 bar indicated mean effective pressure) operating condition was introduced with a single injection at -23.6° ATDC. The data were acquired during the mixture preparation period from near the start of injection (-22.5° ATDC) until the crank angle where the start of high-temperature heat release normally occurs (-5° ATDC). In the present study the measured in-cylinder images are compared with a fully resolved three-dimensional CFD model, namely KIVA3V-RANS simulations. The impacts of computational grid resolution and of the flow initialization method are discussed as they pertain to the mixture preparation process. The simulation results indicate that a fine grid resolution is required to capture the nominal spray penetration in the experiments, however coarse and fine grids are shown to yield similar overall fuel distributions at the start of combustion, which is ultimately important for predicting combustion characteristics and engine-out emissions. It was found that the simulations do an excellent job at reproducing the experimental observation that the majority of the incomplete combustion products (UHC and CO) stem from overly lean mixtures at the start of combustion, which reside in the upstream portions of the fuel spray, near the injector. Additionally, plume-to-plume variability seen in the experiments is explained by possible irregularities in the fuel injector resulting in slightly different fueling rates per injector hole.