I realize this is a very old question, but I guess no harm in me adding my two pennies even after a year.
I agree that it's not good practice writing letters for oneself, but, a lot of bad things happen in academia and this is one of them. It happened to me, and that with a good and dedicated professor: it was just a set of circumstances.
While I asked this person for a letter a few months in advance and he agreed to write it when I apply somewhere, when I stumbled upon my (then, potential, now, current) supervisors, realized we want to work together, there was about a week left to apply for the grant. When I asked for a recommendation, the professor was on a conference / trip and could not find the time to do it himself. He asked me for a "draft" to see what points the letter should cover which he actually did modify later. And I have a feeling he would have preferred to write it than to just modify my writing, but, hey, circumstances.
So, the points I was focusing on when I was writing the letter:
For each letter-writer I contacted, I had a specific purpose in mind. Each of those people could attest to a different set of skills and give a different view of me.
This particular letter was to attest how I have diverse interests (the professor had me on some small extra-curriculum classes), and how I'm good with working with students and explaining stuff (I did some T.A.-like work for him).
So, basically, I wrote about the experiences, facts and results that made me want to ask this person for a letter in the first place. Shortly: Keep to the point and don't digress too much.
Support your statements by facts.
(I guess) nobody wants a read list of synonyms for "awesome". A good rule might be: situation (in which the professor was working with you), result (of your working in that situation), conclusions and reasoning (about your ability, based on the situation).
Try to put the conclusions in context with what you are applying for.
If you are writing a letter attesting (among other things) that you have diverse interests. Maybe you're applying something that is slightly different from your current/previous works, and having diverse interests and an aptitude for learning is certainly a strong point.
Make sure your language is flawless.
Keep to a structure. (not like my bullet points)
A good one might be: firstly, introduce the professor. Shortly list all or some of the situations which make that professor a relevant and good choice for a reference. Secondly, for the "meat", expand on the reasoning for the recommendation. Lastly, summarize the good points and their relevance to the position you are applying for, and include a sentence explicitly saying that you the professor would recommend you. Something, maybe, like Based on my experience with Mister X., I would wholeheartedly recommend X for the position Y.
Don't write a novel. One page should be quite enough. (again, not like my answer :) )
I think this advice should not only help you to write a passable / good recommendation letter, but also increase the chances of the professor actually reading and reviewing what you wrote instead of just signing it. If everything is written concise and to the point, the professor (even the one with very little free time), might be more inclined to change things, because he can identify faster what he disagrees with, what he maybe wants to expand, add, or omit.