The motivation here is crystal clear: e-mail addresses function as (essentially) UIDs on the web as a whole, and someone like Google has billions of these e-mail addresses laying around. But since people started managing parts of their social graphs on Facebook instead of through (say) Gmail, Google's edge graph is starting to get stale. Facebook is terrified of the idea that Google could connect all the Gmail-to-Gmail edges without finding every one of those users.
In order for Google to refresh that graph without e-mail addresses, they'd need every involved user to export their Facebook UID out of Facebook - if I didn't do this myself, then despite the fact that (say) 30 of my Facebook friends are on Gmail, even if every one of them exported from Facebook to Google, Google could never match my Facebook UID to my Google account.
If they were able to match e-mail addresses, they could probably reconstruct a very healthy proportion of the edges (almost every edge that has a Gmail user at both vertices) even if only a small proportion of initial users seeded the graph by exporting their own data; further, when Google got an edge (suppose my friend exported his data, and my Gmail address matched), they could follow up with that edge, knowing that they are also Facebook users, and suggest that they export their data as well, which would lead to blisteringly fast filling out of that tree.
To anyone that doesn't already have a huge userbase, this is not a big deal, but to anyone that does, it's the difference between having effectively zero access to the social graph and being able to reconstruct your users intersection with it with almost perfect resolution.