I disagree with your comment. As I have answered in more detail here: http://www.wikitree.com/g2g/223957/recording-triangulation?show=224549#a224549 we're talking not about shared 3cM segment as evidence of a relationship.
We're talking about a triangulated group where all DNA cousins of this TG share one ancestral segment (eg. chromosome 12, 105-113 mBp). It's normal that not everyone in this ancestral segment has the same start and end position or all the matching segments have the same length.
Hugh is pointing out that one of the DNA cousins in the group has a 3cM segment with one of other DNA cousin. He hasn't given any explicit information about how long the other DNA cousins matching segments are but my guess is that some of them are larger than this specific one.
No matter what the person in question matches himself and four other DNA cousins on this ancestral segment. That is the only proof needed to show a relationship.
A paper trail on the other is always the ideal but in some case impossible (adoptions and NPE come to my mind). So when you write:
I wouldn't take a single shared 3 cM segment as evidence of a relationship - particularly when you do not have a paper trail.
then it's not true that the missing paper trail is making the relationship invalid. Otherwise adoptees would always have invalid relationships until they finally find a paper trail.
DNA genealogy doesn't work that way, it's the combination and the best of both, DNA and Genealogy: Proving your family tree through triangulation