Yes, there is some similarity in expressive power of Lisp macros and Haskell monads (edit: raw Lisp macros are even more powerful, since they may glue and transform anything to anything). However, there is a huge difference in the essence: Lisp macros work on the syntactic level, monads compose on the semantics level.
Lisp macros are not bound by types, so they can be combined irregardless of a program semantics, potentially producing any kind of incorrect code as the result. Such errors get especially tricky with several layers of quasiquotation.
In Haskell, the program meaning is partially encoded into types, so the compiler can understand your intention much better and assists you in programming it right. Also, getting some kind of inter-part optimization is much easier in Haskell: you know exactly what kind of pieces you are gluing up.
