4 responses
Hmmm ... the problem is that "suitable for representaiton in the relational model" is pretty fuzzily defined.  All data can be normalized into relations, if I recall Codd correctly, even if that doesn't mean it can be usefully processed with an existing RDBMS.  Would semistructured data suddenly become structured if there were a great advance in RDBMS' ability to shred and query XML trees or RDF graphs into tables?

Maybe if you said "easily represented in the relational model in a way that could be efficiently implemented in SQL 99" or something?
Yeah, there is no clear metric for how much more useful a particular data model is over another.  It's more a modeling insight than an exact science.  Generally, domains which are subject to unpredictable change are more easily lumped in this category. 

At some point the process of 'normalization' into binary, two-valued relations becomes more bastardization than synthesis (and the result ends up having no use), and this is often the case when you have much more relations than 'entities'.

Also, I'm speaking to the relational model itself not any specific implementation of it (RDBMS').  If the state of RDBMS were to advance to the point where they could accomodate the 'vanilla' relational model, XML content, and an RDF model (simultaneously) that would be a testament to the implementation leveraging less restrictive data models not an advancement of the relational model itself to accomodate semi-structured data.

Any discussion of XML's role as "semi-structured" data should at least footnote this excellent Michael SM quote: http://www.ibiblio.org/xml/quotes2006.html#quote2006April10
Michael SM's quote is garbage. How this nonsense gets published in the ACM I don't know (that issue of ACM query was however generally good for the mostpart). Just consider his opening line:

"In XML, data can have an elaborate and intricate structure that is significantly richer and more complex than a table of rows and columns. "

How anyone could say that XML - with its limited network nature and inherent query bias - is richer than the relational model is beyond me, and indeed to anyone who is knowledgable of relational theory.

I am certainly not saying there is no place for XML but for pete's sake, that particular quote is dire.