I’m going to speak in my native tongue, which is T-Carrier. Concepts are the same for E-Carrier.
You cannot derive timing from the line with a DS-3 (not in the way it is done with DS-1s anyway). Briefly, T-1s can be run through a system called a Digital Access and Cross-connect System (DACS) that can actually terminate an incoming DS-1 frame, move the component DS-0s across a switching fabric, and regenerate a "new" DS-1 frame on the output. This output is referenced to a Building Integrated Timing System (BITS) clock (this is how "line" or "telco" timing is introduced into a circuit). This is not possible (or at least not practical) with DS-3s, which are entirely asynchronous (or, more accurately, plesiochronous). Unlike DS-1s that have fixed time slots (DS-0s) that never move within the frame structure and are thus "synchronous," DS-3s use bit-stuffing to allow for slop in their time slots (DS-1s) and are thus plesiochronous. This is because trying to get a bunch of T-1, from a bunch of different places, to be at exactly the right place at exactly the right time (slot) would be like herding cats - and would require huge, expensive, high-latency buffers.
From discussions I've had with design engineers who work for DACS manufacturers, DS-3 DACS are simply designed as pass-through devices - whatever comes in goes out (within ANSI spec - which is about +/- 400 bps). Thus, the output is not referenced to the BITS and is not "timed" by telco.
Of course, there are always exceptions to any rule. If your telco actually terminated your incoming DS-3 and demuxed it with a real mux instead of a DACS, it could remux it with a piece of equipment tied to the BITS. This would appear to be "line" timing to the customer. We almost went that expensive route with a customer once but they eventually accepted the idea of using their own stratum 3 oscillator on one end to time the circuit and recovered clock on the other. It sure beat buying the telco a pair of muxes and paying them for the space, cooling, power, maintenance, etc.
I sure wish Cisco would add a BNC clock input to their high-end routers or modules but most folks don't really need that kind of clock stability. The routers do fine with their own internal clocks and the entire telco network was designed to accommodate the slop they put out. Even SONET eventually compromised and introduced "byte" stuffing on the tributary side to allow for slop coming in - all the while maintaining the strict clock requirements on the transport side.
