In SUSANA E.U. project a rather broad CFD benchmarking exercise was performed encompassing a number of CFD codes, a diversity of turbulence models... It is concluded that the global agreement is good. But in this particular situation, the experimental data to compare with were known to the modelers. In performing, this exercise, the present authors explored the influence of some modeling choices which may have a significant impact on the results (apart from the traditional convergence testing and mass conservation) especially in the situation where little relevant data are available. The configuration investigated is geometrically simple: a vertical round hydrogen jet in a square box. Nevertheless, modeling aspects like the representation of the source and of the boundary conditions have a rather strong influence on the final results as illustrated in this communication. In other words, the difficulties may not be so much in the intrinsic capabilities of the code (which SUSANA tends to show) but more in the physical representation the modelers have. Even in the specific situation addressed in this communication, although looking simple, it may not be so obvious to grasp correctly the leading physical processes.