If you treated skill levels as a continuum, there would be little difference between books of quality 14, 15 or 16. But if you assume that, beyond being a mere game mechanic, Mythic Europe writers have some awareness of the existence of discrete skill thresholds, the optimization metagame does carry over to the perceived value of individual books.
That means that, if your target audience is raw novices, a L5 Q15 summa is worth significantly more than a L5 Q14, while a L5 Q16 is only a marginal improvement.
Beyond merely maximizing Level and Quality, a writer needs to know his target audience, the goal being to let them reach a threshold as efficiently as possible:
Initial skill level
Target skill level
Expected study time
Reader's aptitude (Poor Student, Book Learner, Study Bonus, Affinities...)
While Roots should be at least L5 Q15 to fulfill their role, I am of the opinion that standard texts should benefit a wider audience than the "average magus", and thus should be useful to both poor and exceptional students. In fact, with L+Q=20, the guidelines in Covenants indicate that this would be a rather poor book, maybe justifying their relative cheapness ("Yes, of course I will let you copy it."), but probably not their status.
If you increase their quality to 18, they become useful even to magi with the Poor Student flaw. However, exceptional students will probably find them tedious ("Why did he waste so much time on something so obvious."). We can up the level a couple notches to L7, catering to all but the most exceptional (Book Learner + Affinity) reader, letting them squeeze a little extra out of the book.
L7 Q18 give a total of 25 points, in the "vain summa" range: neither treasured nor worthless, they can be available from many sources, magi won't mind letting you copy them, give them away when they're through with them... On the other hand, I don't think they should really be perceived as targetting level 7 (they would compare poorly with other level 7 summae), but rather as targetting level 5, with some extra insight for advanced students. That would alleviate the "vain summa" stigma.
I thus think that L7 Q18, with individual variation, is a fair guideline for "The Roots of (Art)", accepted standard primers.
For higher level books, such as the Branches, there certainly is going to be a lot more individual variation: the level of knowledge expected of the target audience, as well as the number of seasons the book is supposed to be studied for can vary widely. On the other hand, tractati become increasingly attractive as level increase.
I have done a quick comparison between skills and arts based on the guidelines page 8 and the amount of experience needed:
basic skill 2 -> art 5
moderate skill 3 -> art 7
skilled 6 -> art 14
very skilled 9 -> art 20
Note that the minimum skill level (2) required for training someone (see page 164) exactly matches (15xp) the minimum art level (5) for properly training an apprentice. I don't think that's a coincidence.
So, an excellent (Q+L=35) summa can take you from 7 to 14 in one year, with a bit of wiggle room for gifted readers (L15 Q20). Alternatively, Q20 L15 can take you from 14 to 20 in about two years of study, while still being better than most if not all tractati.
Anything in that range could qualify as a Branch, depending on your tastes.
Incidentally, for people who want to play with numbers in a spreadsheet:
xp = art*(art+1)/2
xp = 5ability(ability+1)/2
art = (square_root(8xp+1)-1)/2
ability = (square_root(8xp/5+1)-1)/2
I haven't really investigated numbers for ability summae.