Matt,
I guess it isn't obvious that well organized is better, so I'll try to answer your question in another way. Suppose a Beethoven playlist is mistakenly deleted, life having moments of imperfection (and me having a small child who likes keyboards). If the song files are organized in some cases by composer, some by artist, and some by album, then how long will it take me to recreate that playlist, if I ever can? With my present collection size I will have to wade through about 2000 Music Folder root files, then I would have to look within SUSPECTED folders, say "Leonard Bernstein," to see if there was some Beethoven inside. Then many Classical albums contain multiple composers, which would leave me virtually unable to locate the relevant portions except by adding the album to the Library so I can see the file details. The answer to your question seems as self-evident to me as why dictionaries are organized alphabetically: it isn't possible to find things very easily otherwise. And I find myself looking at and fine-tuning song file organization (on the hard drive) frequently. I couldn't begin to calculate the hours I've spent changing the ID data so that it is more accurate and better organized.
I have 943 songs in my Count Basie playlist. Why would I not want to find all those Basie songs contained in subfolders withing the root folder "Count Basie."? I remember quite clearly the huge advance that came with the Mac's heirarchical filing ability in 1984. Now that was actually meaningless and the better way is just to let iTunes jam huge numbers of disparate folders inside of one large folder? Why restrict it to music. Why not have every application dictate the name and location of each file it creates, maybe putting every file on the Mac in one folder called Jumbo?
Should computers adjust to the way people work, or should people be forced to conform to the demands of computers?
It's fine with me if people have iTunes organize their songs, but for iTunes to be able to keep track of the disc location of added tracks is not splitting the atom. Databases, including, for instance, Extensis Portfolio, have been able to do this for many years.