The truncation of diffraction patterns in residual stress determination is often observed for broadened peaks when the 2θ acquisition range is not wide enough. The loss of information effects induced can either be traduced by a bad estimation of the background line, for the methods including a background subtraction, or a restriction of the analysis area for the others. In that borderline case, the results obtained by all methods with theirs specific parameters, developed to estimate the peak localisation are rather distributed in a wide range of stress values. In this paper we propose to review and to test some of the most common methods for stress evaluation (parabola, middle of chord, centred centroïd, asymmetrical pseudo-Voigt fitting). A separate study is made concerning error introduced on the 2θ peak position and on the final stress value estimated. For the parabola method, an analytical expression including some approximations such as the peak shape and its full width at half maximum is then given for the prediction and the correction of these errors. This study is sponsored by PSA PEUGEOT-CITROËN, RENAULT and SNECMA.