While a single T1 is limited to 1.5m, carriers offer higher bandwidths using a "bonded T1" configuration. Essentially it's a number of T1 lines bunched together to act like a single larger pipe, which I'd guess is what you've got. The carriers generally don't let on that there is a difference between this bonded configuration and an actually larger single 5MB pipe. However, most bonded T configurations could only actually achieve the larger claimed bandwidth if you added a number of concurrent transfer rates together. Reason being, and the most likely reason you get a less-than-5m throughput on your transfer tests, is that any individual transfer is limited to traversing one of the T1s in the bundle at a time. If you run multiple simultaneous transfers to different sites you should get roughly the same rate on 4 or 5 transfers at once that you would get on one at a time, as long as their balancing method in their CPE (customer premise equipment, such as their on-site DSU or router) is working properly. They usually configure for a per flow balancing method (think of a flow as a conversation between two hosts on a unique pair of TCP or UDP ports), or less attractively a per IP or block of IP balancing method (less attractive if more than one of you may be conencting to one site, meaning you'd all get stuck sharing one of the T1s in the bundle).
To get past that limitation, you either need "per packet load balancing" turned on in the CPE (usually a bad idea, can cause a number of apps to work erratically), a fractional T3, or some other large single pipe carrier connection.
How's that? :)
