According to Danny Krøger:
But is there any reason for Textmate to change the encoding of the document just because I paste some text, and can I avoid it ?
TextMate will change the encoding if it detects that among the characters you're pasting don't exist in latin1 (which is the right thing to do as it can't possibly which of the various latinN it should use).
Allan, being a die hard utf-8 fan, won't support anything other than latin1 (which is a bit of a pity but I understand him).
That's easy for me to say but converting everything over to utf-8 would be better in the long run.