One possible flaw with this system is that you could have propositional attitudes towards propositional attitudes, ad infinitum. So we should expect an infinite number of classes of affects (beliefs, desires, desires to desire, desires to desire to desire, etc.). I'd like to see the theory discuss whether we observe this; and if not, why not.
I asked Fyfe about this. The system handles a "desire that I have a desire that X" in exactly the same way that it handles any other desire.
The best theory of morality I've ever found is the one invented by Alonzo Fyfe, which he chose to call "desire utilitarianism."
This short e-book (warning: pdf), written by a commenter on Alonzo's blog, describes the theory very well. He also wrote a FAQ.
One great advantage of this theory is that what it describes actually exists even if you prefer to use the word "morality" to mean something else. Even a community of paperclip maximizers may find something in it to be relevant, amazingly enough.