The TripleTermData rule now allows generalized triple terms like <<( "foo" ex:p "bar" )>>. This breaks the possible assumption in SPARQL implementation that the RDF terms in VALUES are always valid RDF. It might even break wider assumptions like "manipulated RDF terms after parsing are always valid".

It might make sense to change TripleTermDataSubject into:

[123]   	TripleTermDataSubject 	  ::=   	iri