Markdown generates invalid markup

The following snippet of text generates invalid markup:

* ABC
    * DEF

    *GHI

The result is (with @e417fd0 and the latest version from PyPI):

<ul>
<li><p>ABC</p>

<ul>
<li>DEF</li>
</ul>

<p><p>*GHI</p></li>
</ul></p>

Note that that second-to-last <p> tag is opened inside the <li> element, but closed outside the <ul>. This breaks my XML parser!

The source markup is not exactly clean (I am not even sure what the correct result would be) and is easily fixed by adding a space before "GHI". This still is an interesting case though, because it is genuine, non-technical user-input that I found in one of my wikis. As I parse the output of markdown2 as XHTML using ElementTree, this broke the rendering of the whole page...

Here is the result of parsing the same snippet with the "regular" markdown library:

<ul>
<li>
<ul>
<li>DEF</li>
</ul>
<p>ABC</p>
<p>*GHI</p>
</li>
</ul>

... and just for fun, here is what happens when GitHub parses it:

• ABC

  • DEF

  *GHI

Reproduce the reported snippet through markdown2 and inspect the generated list markup against the regular Markdown output shown in the issue. Determine a valid XHTML structure for this input, then verify that the resulting output has properly nested and closed elements and can be parsed by ElementTree.