If it is Nahuatl, it should be possible to decipher a lot of it pretty quickly. Nahuatl has very regular morphology which will give repeating patterns that can be used to decipher the letter, assuming they're phonetic. For instance most nouns end in -tl so we should see many of the words in the manuscript ending in a symbol or two symbols which we can assume is tl.

Maybe someone can apply this model: http://pages.cs.wisc.edu/~bsnyder/papers/bsnyder_acl2010.pdf