R5RS gives proposed macro definitions for library forms of syntax:

http://schemers.org/Documents/Standards/R5RS/HTML/r5rs-Z-H-10.html#%_sec_7.3

Which also defines letrec, in a very complicated way, certainly not how I would define it, I would simply use:

(define-syntax letrec2
  (syntax-rules ()
    ((letrec2 ((name val) ...) body bodies ...)
     ((lambda ()
       (define name val) ...
       body bodies ...)))))

As far as I understand the semantics of letrec, which I use very often as a named let. It works in this way, however as I've had my fair share of debates with philosophers who think they can just disprove special relativity or established phonological theories, I know that when you think you have a simple solution to a complex problem, it's probably WRONG. There has got to be some point where this macro does not satify the semantics of letrec else they'd probably have used it.

In this definition, the definitions are local to the body of the letrec, they can refer to each other for mutual recursion, I'm not quite sure what (if any) is wrong.