I don't think that's what is being said. The statement is more like 'any imperative program can be represented by a purely functional representation', which goes beyond simply modeling loops with tail call elimination.