Weird output with *print-circle* t

[vi1]

;; ccl output this: " #1=___ qwe #1# "
 
(setf *print-circle* t)
 
(defun centered-string (n string)
  (let ((w (max 0 (floor (- n (length string) 4) 2))))
    (let ((m (make-string w :initial-element #\_)))
      (format nil " ~a ~a ~a " m string m))))
 
(print (centered-string 14 "qwe"))

[svspire]

Here's a simpler example:

(defun simpler (string)
  (with-standard-io-syntax 
      (let ((*print-circle* t))
        (format nil " ~a ~a " string string))))

(simpler "foo")

? 
" #1=foo #1# "

This appears to be the same issue as described in Issue PRINC-READABLY Writeup where Macintosh Common Lisp is specifically called out for following CLtL2 too literally. Since CCL inherits much of its implementation from MCL, it makes sense that CCL has the same issue.

I'm just surprised that MCL (and CCL) were never changed to adopt the above linked proposal. Lispworks and SBCL both seem to have adopted it, and IMO adopting the proposal produces the behavior that I'd expect.