This question deals with letters of recommendation written for students in the humanities, and in particular for those on the job market or those submitting to dissertation grants and fellowships. The question may also apply to letters of recommendation in the sciences.
On this site and in conversations with current faculty members, I have surmised that letters of recommendation are usually written in a kind of "code" (one post here described these letters as a "second language" you have to learn). Since most letters are invariably positive, I suppose that the code is just a subtle way of signaling the true value of a candidate's work or the true extent of their potential. Those who know the code are able to see through any hyperbolic language and decode a letter-writer's real estimation of their candidate.
A common warning about asking for letters of recommendation is: make sure you ask someone who already knows how to write them, or they could end up doing more harm than good. It makes sense that, if someone doesn't know the code, that person won't be able to communicate whatever it is that committees want to see in the letter.
This is all just my ignorant explanation of what I've observed, however, and I would like to ask for more information. I am especially interested in knowing:
-Is it true that letters of recommendation are written in such a clandestine way?
-If there is a code, what is it like? What kind of signals does it have? How do you know when you are reading about a candidate who really is great, versus a candidate who is just fine, when their letters may contain similar language?
If it turns out I am off base on this, I'd appreciate hearing about that too.