I learned Morse code from cassette tapes sold by the ARRL, in which the characters were at 15 WPM but the characters were spaced at 5 WPM; in other words, the Farnsworth method. The Koch method has the characters and the spacing at the same speed, which is relatively fast. Both the Farnsworth and Koch methods introduce the learner to a few characters at a time.
I would think that there are two considerations involved. First, the trainers want to introduce the learner to the easy characters, the ones with only one or two dits and dahs, earliest. The second consideration is that they would like to teach characters that are easy to make into words, because it is more satisfying for the learner to copy complete words than to copy random groups of characters.
I haven't done a scientific study, but in my opinion it's more important to make words than it is to teach the "easy" characters. Personally I remember that for me, some of the longer characters such as P (·--·) were easier to recognize and memorize than some of the shorter characters, such as N (-·). Of course the International Morse Code is designed around the English language, so that the easier characters are also the some of the most commonly-used characters, so there isn't much of a conflict between the two considerations.
Regardless of the method used to learn Morse Code, I think the best thing is to get through the awkward phase of learning all the characters as quickly as possible. It is far more satisfying to use the code in an actual QSO, even at a slow speed, than it is to listen to a tape or use a computer program.