From a conversation with Mathias Brochhausen:

... there is some mess with report and publication being parallel and journal article being one and publication about investigation being in other.

On quick inspection it looks like the hierarchy should be:

report publication journal article publication about an investigation

With publication about an investigation being a defined class, as otherwise journal article about an investigation would have two asserted parents.

I think the genesis of this was parallel work by OBI and Larry Hunter that wasn't fully integrated.

See also related http://code.google.com/p/information-artifact-ontology/issues/detail?id=33