With function `ucs-insert' (from lisp/international/ucs-tables.el):

  M-x ucs-insert RET 4DF9 RET => "Character U+4DF9 is not yet
supported" (an error message).

With ucs input method:

  U4DF9 => "ä·¹" (in the buffer).

Why does the ucs input method try to insert what it knows is an
unsupported character? In which way it is better to get "ä·¹" than a
helpful error message?

I ask because the behavior is clearly intentional. `utf-8-ccl-encode'
is defined in leim/quail/uni-input.el, and `ucs-input-method' includes
this error handling:

 (if c
    (list c)
  (aset status 0 n)
  (string-to-list (ccl-execute-on-string
                   'utf-8-ccl-encode status "")))

                    /L/e/k/t/u