From a bit of experimentation with 'xtract' it appears that the order of the command-line arguments is important, and thus a use of '-match' must be followed by an '-element' option. This appears to be the source of the error message you receive.

Using just 'xtract' the closest I've gotten so far is:

cat xml.txt | edirect/xtract \
  -pattern Gene-commentary \
  -match 'Gene-commentary_type:1' \
  -element 'Gene-commentary_accession' Seq-interval

You may be able to further anchor the patterns to make the extraction more specific.

Using a good old xslt stylesheet:

<?xml version="1.0" encoding="UTF-8"?>
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform" version="1.0">
  <xsl:output method="text"/>
  <xsl:template match="/">
    <xsl:for-each select="/Entrezgene-Set/Entrezgene/Entrezgene_locus">
      <xsl:for-each select="Gene-commentary[Gene-commentary_type/@value='genomic' and Gene-commentary_type/text()='1']">
        <xsl:variable name="acn">
          <xsl:value-of select="concat('(',Gene-commentary_heading,')',Gene-commentary_accession)"/>
        </xsl:variable>
        <xsl:for-each select="Gene-commentary_seqs/Seq-loc/Seq-loc_int/Seq-interval">
          <xsl:value-of select="concat($acn,':',Seq-interval_from,'-',Seq-interval_to)"/>
          <xsl:text>
</xsl:text>
        </xsl:for-each>
      </xsl:for-each>
    </xsl:for-each>
  </xsl:template>
</xsl:stylesheet>

run:

xsltproc --novalid transform.xsl  "http://www.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=gene&id=672&retmode=xml"

output:

(Reference GRCh38 Primary Assembly)NC_000017:43044294-43125482
(Reference assembly)NG_005905:92500-173688
(Alternate CHM1_1.1)NC_018928:41431850-41513017
(Alternate HuRef)AC_000149:36962662-37043808