Ambiguity regarding xsi:type in OPF 2.0.1

The Relax NG schema in Appendix A has been updated from OPF 2.0. Every instance of the optional "xsi:type" attribute has been excised from the schema. For instance, this is what the "language" element rule used to look like in OPF 2.0:

<define name="DC.language-element" ns="http://purl.org/dc/elements/1.1/">
  <element name="language">
    <ref name="OPF20.optional-id-attribute"/>
    <ref name="OPF20.optional-xsi-type"/>
    <ref name="DC.metadata-common-content"/>
  </element>
</define>

and this is what it looks like now:

<define name="DC.language-element" ns="http://purl.org/dc/elements/1.1/">
  <element name="language">
    <ref name="OPF20.optional-id-attribute"/>

    <ref name="DC.metadata-common-content"/>
  </element>
</define>

There is literally a hole where the "xsi:type" reference used to be. Looking at the page source, you can see that the reference is still actually there, but wrapped in "del" tags (meaning that it was intentionally removed).

BUT the Publication Metadata section still contains this text in a paragraph near the end:

For compatibility with Guidelines for implementing Dublin Core in XML (http://dublincore.org/documents/dc-xml-guidelines/) this specification allows xsi:type attribute for metadata items that can be given using some sort of encoding scheme and xml:lang attribute when an item can be given using human-readable text. Elements that allow xsi:type attribute are identifier, language, date, format and type.

This text was in OPF 2.0, but it now doesn't make sense if the schema doesn't allow the attribute since the schema is also normative.

So which section am I supposed to trust? Was this text just forgotten or what?

[originally posted as a part of this thread on MobileRead]