You can't "remove" the CDATA, but you can achieve the desired output somewhat crudely:
<?xml version="1.0"?>
<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:template match="/">
<Detail>
<xsl:variable name="before" select="substring-before(//Detail,'<div class="heading">')" />
<xsl:variable name="afteropen" select="substring-after(//Detail,'<div class="heading">')" />
<xsl:variable name="body" select="substring-before($afteropen, '</div>')" />
<xsl:variable name="after" select="substring-after($afteropen, '</div>')" />
<xsl:value-of select="concat($before, '<h1>', $body, '</h1>',$after)"
disable-output-escaping="yes" />
</Detail>
</xsl:template>
</xsl:stylesheet>
This will work for the first type of div you're trying to parse and you can follow something similar with the second one. It could be made more generic with some effort.