Rutherford backscattering spectrometry (RBS) is an ion beam based analytical technique giving information on the atomic composition profile of the near surface region of materials. As the physical processes involved in this technique are very well understood, computer codes have been developed providing simple and efficient means of performing quantitative RBS analysis. Moreover, with this technique it is possible, in principle, to perform absolute measurements although the task is not an obvious one. The attainable accuracy in the measurements is limited, among other factors by the uncertainty in the current knowledge of the stopping power of materials for alpha-particles of MeV energy. The shape of an RBS spectrum from which an energy/depth relation and a yield/composition relation has to be extracted is dependent on the stopping power values, and its uncertainty directly affects the attainable accuracy of the analytical information of the spectrum. In this contribution we show evidence of the inaccuracies of the approximated stopping powers in the computer code RUMP for the matrix element Si by direct use of Harwell II standard samples in a RBS experiment, and show how these inaccuracies can affect the interpretation of the results in a RBS analysis.