This report concerns the last edition of the XML Mining Track at INEX 2007. A preceding report has been already published concerning the two preceding editions of the track. We present here the new corpus used for this third phase and briefly describe the models and the results obtained by the different participants.