Hi Chris - the answer is that Mendeley doesn't check the metadata provided in the PDF (because it is usually incomplete and often wrong), but tries to guess the correct metadata from the full-text of the document and checking it against external databases (such as PubMed, CrossRef, arxiv, Google Scholar etc.). However, this is a case where it obviously would make sense to use the PDF's embedded metadata - I just filed an internal ticket for this! Thanks for pointing this out to us. - Victor / Mendeley Team

merci Mendeley pour la réponse (Cc : @neuromancien @jsicot Bcc: @mendeley_com ) - chris dabin