When connected to a circuit, you should use both, (source line) plus a third, network-clock-select, giving priority to the interface in question. This way, you will accomplish two important goals:
1: for reception, the hdlc framer is clocked off the line rather the internal source, that inherently cannot be synchronous. Benefit: no input slips.
2: for trasmission, you are using the same received clock and be confident that thins will be ok at the remote circuit termination. Benefit: no remote slips.
Else, your T1/E1 interface can encounter slips, especially on the new ISR platforms that have a sophisticated clock system.
Now for when you are connecting to a local equipment with passive cabling, you must know if the other side is generating reliable clock. If not, you need to use clock internal and PLL internal source. The other device will take and reuse this clock and everything will working nicely.
This is my nutshell explanation. On CCO, there is a lot of material on the subject.
Hope this helps, Please rate all useful posts!