Recently, a blind CFD benchmark exercise was conductcd by the OECD/NEA (2013-2014) based on an experiment in the PANDA facility at the Paul Schcrrcr Institute (PSI) in Switzerland, investigating the turbulent erosion of a stratified helium rich layer in the upper region of the test vessel by means of a vertical air-helium jet impinging from below. In addition to the 'classical' pointwise measurements available for similar experiments conductcd in the past, significant additional efforts were spent on the experimental characterization of the underlying flow field and turbulent quantities by means of particle image vclocimctry (PIV) for the benchmark. This data is well suited for a detailed assessment of the driving jet flow and its interaction with the stratified layer. Both are essential in order to avoid elimination of different errors, which is possible if validation is performed in a global manner. Different impacts on the simulation results, in particular on the jet profile and on the mixing progress, arc discussed in this paper. A systematic validation is canicd out based on measured and derived quantities. It is identified that e.g. the mesh resolution in the jet and mixing zone has only a minor impact, while small changes in turbulence modeling strategy or the chosen model constants, like Sc:. significantly affect the simulation results.