To be a longa the first note of the phrase (case 1) must be the pitch of the modal final. The second note of a multi-note syllable (case 2) is only a long if it is not preceded or followed by another one of the five exceptions. A single plicated note (case 3) is a longa when the note plicated is itself written as a longa. If it is separately written it has the value of an imperfect longa, but if it is ligated it could be a perfect longa or even a four-unit longa. (The actual value would depend on neume shape and notational context.) Also, a pair of plicated, ligated notes with the written value of two breves could have the value of breve-longa if they are followed or preceded by a longa.From p. 65 of The Sound of Medieval Song, by Timothy J. McGee.
Believe it or not, this is a book that I've wanted to read for the last five years. (Thanks to Prof. Jesse Rodin for describing it to me back in 2009.) Don't ask me to translate the above; I sort of understand what's going on - sort of - but I would have to review a lot of terminology (and draw diagrams) to explain it to someone who is not familiar with early music notation.
I can also report the odd circumstance that I have met Prof. McGee, long ago, and will be dropping him a note to say hi.