In emacs 21.3, lisp/language/cyrillic.el has a defvar 
cyrillic-koi8-r-decode-table that only covers the letters of KOI8-R (and 
maps everything else to ISO 8859-1).

It'd be trival to make it use the actual table from RFC 1489 (which I 
think is the right thing to do).

Attaching proposed patch (in utf-8).

- Gustav