The complaint that the dots and dashes ambiguously decode into billions of letter combinations ignores the fact that one could probably safely assume that the final message is supposed to be sensible English. Instead of attacking the dots and dashes and trying to form words, one should start with a word list, convert its entries into Morse, and see which words could potentially start the decoded message.
Now, I don't expect non-programmers to be able to do this, but with the length of a Marathon, the coders on the various teams might be able to cook something up that helps. For example, here is a web-interface backed by a Perl CGI script that helps. It does not solve the whole message in one swoop, but with some human interaction and intelligence manipulating it.
I use the /usr/share/dict/words.
Of course, I understand that actually transcribing the original Morse message from tree stump to computer screen is no trivial task; but not undoable.