In this paper, we address the problem of testing radiated emissions generated by rolling stock, when carried out on actual railway sites, as prescribed by the standard EN 50121. The idea of assessing infrastructure impact on test results is presented here by means of an electromagnetic model of railway sites. In the first part,modeling tools are introduced together with some results from the experimental validation. These tools are then applied to actual railway lines proving the importance of site resonances and the ambiguities in the application of the standard EN 50121 in the industrial domain. After pointing out the main difficulties in the simulation of actual railway sites, we present a feasibility study of an alternative procedure for the modeling of a site, based on magnetic field measurements and the solution of an inverse problem with no a priori information about the test site configuration.