The Life Cycle Assessment (LCA) methodology is commonly employed to assess the performances of the different uses of biomass, which is supposed to play an important role in both mitigating climate change and reducing dependence on fossil fuels. The range of LCA results is however very wide and prevents from having reliable and accurate results for policymaking. In order to better understand such discrepancies in results, this work reviews the LCA methodology and issues, and carries a sensitivity analysis on every LCA parameter, either assumption or data. Nonspecific methodological elements appear to be of major importance for LCA results: allocation rule influence is quantified to range from 48 to 82%, and system boundaries one from 0 to 93%. Input data play also an important role in results discrepancy, especially for GHG emissions. Local specificities can influence the final results by 53 to 118% and N2O uncertainties by 31 to 67%. This calls for agreements about LCA methodology depending on its application, limitations of geographical coverages and integration of the best scientific knowledge about N2O dynamics. Finally the soil integration shortcoming is pointed out. Variations of soil carbon quantities due to land use changes can be very important and overwhelm GHG savings induced by biofuels use. This parameter is however lacking in LCA methodology applied to biomass-based products.