XSLT character-escaping query

If I understand you right, you want text nodes to come out as literal text (disable-output-escaping="yes"), but the rest of the transformation should work normally (<bold> to <b> etc.)

Template modes can help:

<xsl:stylesheet 
  version="1.0" 
  xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
>
  <xsl:output method="xml" encoding="utf-8" omit-xml-declaration="yes" />

  <xsl:template match="paragraph">
    <p>
      <xsl:apply-templates mode="literal" />
    </p>
  </xsl:template>

  <!-- literal templates (invoked in literal mode) -->
  <xsl:template match="bold" mode="literal">
    <b><xsl:apply-templates mode="literal"/></b>
  </xsl:template>
  <xsl:template match="italic" mode="literal">
    <i><xsl:apply-templates mode="literal"/></i>
  </xsl:template>
  <xsl:template match="text()" mode="literal">
    <xsl:value-of select="." disable-output-escaping="yes" />
  </xsl:template>

  <!-- normal templates (invoked when you don't use a template mode) -->
  <xsl:template match="bold">
    <b><xsl:apply-templates /></b>
  </xsl:template>
  <xsl:template match="italic">
    <i><xsl:apply-templates /></i>
  </xsl:template>

</xsl:stylesheet>

Had a similar problem transforming ATOM XML where the content of the posts contained escaped HTML tags. disable-output-escaping seems to work in most browsers but not Firefox. So as a work around, in my XSLT I added the class="renderhtml" attribute to all problematic output nodes, as a hint to javascript. Then added the following code after the transform.

// work around for XSLT disable-output-escaping="yes" not working in Firefox
if (navigator.userAgent.indexOf("Firefox")!=-1) {
  // get all "renderhtml" hints (HTML escaped within CDATA/text) nodes, then unescape and render HTML code
  var nodes = document.getElementsByClassName("renderhtml");
  for (var i = nodes.length - 1; i >= 0; i--) {
    nodes[i].innerHTML = nodes[i].textContent;
  }
}