You are seeing the result of XInclude plus cleanup on a trivial XML document from MT's own site --

The pipeline you ran this through to get here (showcase/cleanIncl) depends crucially on Markup's use of a full-featured infoset bus, which makes validation and type information available cleanly and easily to pipeline components. You can see the pipeline description which was compiled to produce this pipeline for the server to run.

The first relative URI
XIncluded elt with relative URI

What was immediately above before XIncluding was:

<a href="foo.xml">The first relative URI</a><br/>
 <xi:include href="subdir/included.xml" xpointer="element(/1/1)"/>

After XIncluding it looked like this:

<a href="foo.xml">The first relative URI</a><br/>
 <a href="baz.xml" xmlns="http://www.w3.org/1999/xhtml"
    target="foo" xml:base=".../subdir/included.xml">XIncluded
elt with relative URI</a>

And finally, after base cleanup and absolutization, as you can confirm by hovering over the links above and/or doing 'View Source':

<a href=".../foo.xml">The first relative URI</a><br/>
 <a href=".../subdir/baz.xml" xmlns="http://www.w3.org/1999/xhtml"
    target="foo">XIncluded elt with relative URI</a>

Running your own data through this pipeline

Replace http://www.markup.co.uk:8888/yourIncl.xml in the URL you're fetching with a full URL (http: or https:) for the page you want to process (presumably containing some xi:include elements),and replace http://www.markup.co.uk:8888/yourPostInclSchema.xsd in the URL you're fetching with a full URL (http: or https:) for a schema document corresponding to the schema you want to validate with. Whatever it successfully identifies as an instance of anyURI will be absolutized with the correct base URI.