Due to the wide spectrum of turbulent and chemical length- and time scales occurring in a HSDI diesel engine, capturing the correct physics and chemistry underlying combustion poses a tremendous modeling challenge. The processes related to the two-phase flow in a DI diesel engine add even more complexity to the total modeling effort. The Representative Interactive Flamelet (RIF) model has gained widespread attention owing to its ability of correctly describing ignition, combustion and pollutant formation phenomena. This is achieved by incorporating very detailed chemistry for the gas phase as well as the soot particle growth and oxidation, without imposing any significant computational penalty. The model, which is based on the laminar flamelet concept, treats a turbulent flame as an ensemble of thin, locally one-dimensional flame structures, whose chemistry is fast. A potential explanation for the significant underprediction of part load soot observed in previous studies applying the model is the neglect of wall heat losses in the flamelet chemistry model. By introducing an additional source term in the flamelet temperature equation, directly coupled to the wall heat transfer predicted by the CFD-code, flamelets exposed to walls are assigned heat losses of various magnitudes. Results using the model in three-dimensional simulations of the combustion process in a small-bore direct injection diesel engine indicate that the experimentally observed emissions of soot may have their origin in flame quenching at the relatively cold combustion chamber walls.