You're absolutely right, the sample document only successfully parses a single author -- an earlier version of the metadata parsing algorithm was designed to actually detect multiple authors and affiliations, but it was found to be far too unreliable, and often produced very poor results. The current approach is to favour detecting one
author correctly, rather than several incorrectly.
The main issue at play here is that formatting and presentation of authors and affiliation information is extremely
non-standard -- this often varies widely among submissions to a single journal, for example, and even among submissions from the same author(s). Using article templates is one way in which some journals and presses have addressed this situation, and while L8X doesn't make use of document templates at present, it is one area we have discussed using more fully.
The issues you're seeing are also key to our decision to integrate L8X into the PKP framework
, including OJS, since this will place detection and editing of authors directly in the context of the journal, conference, monograph, etc. I realize it's frustrating right now, but we're quite optimistic that once people see this functionality in-place, it'll be both more understandable and easier to use.