Unboiling letters of recommendation

Laboratory work, especially demographic laboratory work, is labor intensive. (Perhaps that is why it is called a laboratory?) I couldn't do the experiments I do without lots of reliable helpers. At the Institute where I work, and at most universities I've seen, much of this labor comes from eager young undergraduate students. They work hard (if carefully chosen), want to learn and don't cost a great deal.

This is the last week I will employ most of my wonderful mob of students, and so today I am writing letters of recommendation for almost all of them. I've gotten to know most of them quite well, and I like all of them, so it shouldn't be such a hard task. But of course I have found a way to make it hard. I feel I owe it to them to write a memorable individual letter for each person, not a formulaic boilerplate with a few details changed. And while their personalities are very different, many of the positive things I can say about them in these letters are all the same, leading to boilerplatedness. I suppose the need for creativity within constraints such as these should be taken as an artistic challenge.