In mIRC v6.35, if you have set your window font to use eg. a Greek script/codepage and you then use $utfencode(text) in that window, it will perform the conversion using the Greek codepage. Resetting the fonts also reset your script/codepage for that font, so that is probably why the issue resolved itself.
Also in mIRC v6.35, the characters 128-256 will appear differently to users on IRC depending on their system language or their selected font script in the a window. That means it is not reliable to use this range of characters on IRC - there is no guarantee that users will see the same characters.
That is why mIRC v7.x now uses UTF-8 for all text, to ensure that everyone, everywhere, consistently sees the same characters. In mIRC v7.x, all text is Unicode text. When you use $utfencode() or $utfdecode(), it will convert to/from Unicode/UTF-8. The Unicode/UTF-8 conversion routines check (on a per character basis) if text is already in UTF-8 format and will prevent it from being double-encoded.
