Maybe they're syllabic (which would be reasonable, given the number of distinct symbols), and the language uses repetition of a syllable as a case marker, or just as a stylistic flourish? Perhaps it's used for emphasizing standalone nouns: Scis-issors-ors (scissors), vehic-icle-ickle (vehicle), o-sho-cean-shan (ocean). Likewise, if you're carving it in a building, why not write 'museum' as MVSEVM?
I know it's a conlang, and Serafini is probably just putting us on, but the ancient Mayan script does some pretty remarkable calligraphic somersaults, too.
If I were doing something like this, I would assign new glyphs to each letter (as not to break the image of a fantasy world), and then encode the whole thing with a one-time pad. Doing so can give you all sorts of weird-looking words that have actual value.
Does anyone know how many unique glyphs there are in his book? For that matter, how many are there--including both uppercase and lowercase--in French (I don't know how many accented characters there are)?
And if there are still too many to be such an arrangement, it is possible that each letter was assigned multiple sets (a set being uppercase and lowercase) of glyphs and picking a random number between 0 and twice the total number of letters in French, counting accented letters as unique letters (manually picking the proper case).
I'm not saying that he did this, but it's what I would do.
It's been a few years since I saw the book, but the total number of glyphs seemed a lot closer to (for example) the number of hiragana + katakana than the number of kanji.
(Also, I don't know if you picked just French at random, but Serafini is Italian.)
