My understanding is that book publishers didn’t want the headache of knowing which printers were licensed for specific fonts, so they got a specific carveout for the shape of letters. So when they sell a book, they don’t have to worry about whether the printing company was licensed to resell the books they printed for the publisher.
The “you can have a copyright on the font software” aspect seems to be based on the idea that a particular font file is just as copyrightable as, say, a computer file of a company’s employee handbook describing uncopyrightable methods and systems ( https://www.copyright.gov/circs/circ33.pdf ).
If I pay Microsoft for a cool font, I don’t have to pay extra to print and distribute documents using it, but I can’t email the font file to my friends without paying Microsoft more. It’s not clear to me how that interacts with fonts embedded in PDF files, but I believe some companies charge more to license fonts so they can be embedded in PDFs.
As far as creating a font file based on the physical shape if letters: you would end up with a file similar to the original, but I doubt it would be bit-for-bit identical. So you would have a reasonable explanation of how you got the similar file without violating anybody’s copyright.