By the way, I looked inside the generated output of this my post and it looks alright for my test UTF-8 phrase. How to achieve this?
Ok, the problem is solved. To get right UTF-8 in output, text must be non escaped. E.g. if the XHTML source was:
then it must be:
<h1><h:outputText value="УТФ" escape="false"/></h1>
Now I have got another question: is there elegant way to escape only <>&" but not the every char except of latin?