I suppose that's the point, it's not just Unicode being quirky, imagine trying to publish a paper on language X and its unique, dead, script. It'd be harder now in the 21st century without it being in Unicode than it would've been in the early 20th (without Unicode existing at all). What would you do? Append an image and refer to the characters numerically?
Probably, preferably as a vector file, and then reference it with latex (or whatever you're typesetting your paper with) so it shows up as part of the rendered document. I.e. same way you include non Unicode items in your paper.
