The reliability of ray tracing simulations is strongly dependent on the accuracy of the input data such as the bidirectional reflectance distribution function (BRDF). Software developers offer the possibility to implement BRDF data in different ways, ranging from simple predefined functions to detailed tabulated data. The impact of the accuracy of the implemented reflectance model on ray tracing simulations has been investigated. A light-emitting diode device including a frequently employed diffuse reflector [microcellular polyethylene terephthalate (MCPET)] was constructed. The luminous intensity distribution (LID) and luminance distribution from a specific viewpoint were measured with a near-field goniophotometer. Both distributions were also simulated by use of ray tracing software. Three different reflection models of MCPET were introduced, varying in complexity: a diffuse model, a diffuse/specular model, and a model containing tabulated BRDF data. A good agreement between the measured and simulated LID was found irrespective of the applied model. However, the luminance distributions only corresponded when the most accurate BRDF model was applied. This proves that even for diffuse reflective materials, a simple BRDF model may only be employed for simulations of the LID; for evaluation of luminance distributions, more complex models are needed.