You could try the “Crossref Lookup” script that can be installed via DEVONthink’s script menu (see the “More Scripts…” menu entry). With an article title selected in a PDF, this script attempts to fetch some bibliographic metadata & the article’s formatted citation from CrossRef.
Alternatively, if your PDF contains annotations, you could also try my “DEVONthink Notes from PDF Annotations” script. This script takes a DOI from a PDF record (or extracts it from the PDF’s metadata or its first page) and fetches bibliographic metadata as well as optionally the article’s formatted citation & BibTeX data.
Both options would give you bibliographic metadata for a PDF in DEVONthink. However, this requires that your PDF article has a DOI (digital object identifier).
In the Bookends import dialog, make sure you keep the first two checkboxes (“Move/Copy To Attachment Folder” & “Rename: …”) unchecked and only check the third checkbox (“Attach to a new reference”). Then, Bookends will just link to the original file and won’t move it.
However, there’s one caveat: You should also make sure that a PDF with the same name doesn’t exist already in your Bookends Attachments folder. If there already exists a PDF file with the same name, Bookends will rename the imported PDF and append its own ID to the name. This would render the file inaccessible in DEVONthink.