Previous studies have established the efficiency of training programs to help users improve on binaural auditory localization when using a non-individual head-related transfer function. This paper reports the results of an experiment where participants trained with such a program, though they failed to improve. Interestingly, the same training program had successfully been used in a previous study. After a brief description of the program and the experiment protocol, a comparative analysis of the results of both studies is provided. The discussion then focuses on the differences between these studies in an attempt to better understand what caused the training to fail.