Um, yea, bfd
A single tower can host potentially thousands of concurrent calls. A Femtocell balks at more than 4. Also, at least with AT&T, those femtocells are locked down to only approved devices, and are not open to the public in general, but even on Verizon and Sprint, the range of the femtocell rarely penetrates the walls of the dwellign. (V advertises only 40 feet, ouch!)
And that backhaul? In a large number of cases, that back haul is owned by them anyway. My AT&T cell is on an AT&T line... However, free? 1: the backhaul from towers are their own internal network lines. They don't pay carrier fees on them, only the cost of the physical cable and some power,and some routing muscle. the Femtocells? They may or may not be on their own networks, and incurr 3rd party carrier fees to cross communicate to other provider IP networks, eating external, not internal, bandwidth. Also, the back end equipment to handle those units (which are essentially VoIP extensions in their network), is a completely separate infrastructure, and extremely costly to both design, implement, and keep in sync with the real towers.
Also, why have a femtocell if you had good coverage? What this meas is that 300,000 $50 femtocells are covering what is probably 500,000 underserved or not served at all homes. $15m to add 500,000 users? DEAL! (oh, yea, those users, most of them, paid $100-300 for that femtocell, so lets call it a profit party!) 3 or 4 towers can cost that to deploy, if not more. Are we really surprised this is popular with Providers for underserved areas? In served areas, the per-user cost of a femtocell likely exceeds the per-user cost of a node on a tower, all factors included, and that's why they're making efforts to restrict their use by GPS signal and mapping tools.