Noise effects induced by elongation have to be considered globally in a multi-channel tomographic reconstruction analysis. Such an analysis allows a fine estimation of performance and to compare launching options. We developed a modal LGS tomographic code based on pseudo-analytical matrix formalism. This modal analysis is computing and memory intensive. Therefore we limit the simulation to downscaled cases with diameters ranging from D = 4 to 21 m. We also limit the number of turbulent layers to a few equivalent layers. Including spot elongation and Rayleigh fratricide effect, when considering a global tomographic reconstruction, edge launching gives lightly better performance and limits the risk of possible underestimation of fratricide scatter.