If you are not at an R1 or elite liberal arts college, the ranking needs to be "top 2%" to carry much weight. Having said that, I am suspect of these rankings; they are obviously subjective, and the incentive for the letter writer is to get their student into the best program possible. The letter writer pays no penalty if they oversell a student, so they inflate the ranking. When I see "Top 1%", I discount that to 10%; "Top 10%" is 25%; and "Top 20%" is 50%. "One of the best" means that there are several individuals who were better. I have read tens of thousands of reference letters, and have seen fewer than 50 that actually say anything objectively negative about an individual, and only a very few that say that the individual would not be a good scientist.
Having been at this for many years, I have the benefit of seeing letters from the same advisors. When I see a letter from a prolific trainer of young scientists like Joan Steitz, I can go back to letters she wrote over the past several years to see how the current letter compares with the past letters. While all these letters are self-plagarized to varying degrees, the best ones comment on specific aspects that are important to us, and differentiate among their advisees by their level of praise and use of modifiers (best, agile, rigorous, etc). At the other end of the spectrum are overworked (or lazy) advisors, who basically use the Microsoft Word find and replace function to recycle letters year after year. I saw that earlier this year when the letters for two different applicants were verbatim, except for the names of the applicants. (Fortunately, they were the same gender, so the pronouns did not have to be changed.) If the letters were to be believed, the individuals worked on identical projects, possessed the same characteristics, had the same potential for a career in science, and were "among the very best" the faculty member had trained.
As someone who has to write quite a few letters for residency applications every year, I understand the allure of writing a generic letter, and also of overselling an individual's abilities. However, I know that my audience is small (maybe a dozen medicine program directors, 5 or 6 pediatrics, 4 neurology, etc.), and they will be reviewing several students from my program every year. If all the letters are the same, they become worthless. Additionally, if I oversell a candidate this year and he struggles in residency, that will be fresh on the residency program director's mind when they read my letters for next year's applicants. I owe it to my students to put the effort into individualizing the letters and to be honest about their abilities. Therefore, I get annoyed when I see fairly useless letters from PI's who only have to write 1 or 2 letters a year to MD-PhD programs.
