The reporting on our work has been of mixed quality. Check out our actual blog-post here:
http://webpolicy.org/2013/12/23/metaphone-the-nsas-got-your-... for more accurate details.
We used data collected from voluntary users' phone logs as our phone number data set. This means that if Joe called number X a few weeks ago and then decided to participate in our study, number X was in our database. We then used a couple techniques too see how well we could identify who number X belonged to.
We didn't put together actual profiles of users, though that is a possible next step. However, I think it is clear that putting together profiles of users is possible given how easy it is to identify who you are calling and receiving calls from.