110709-1138 EDT
ken9876:
I do not think shields are a problem for a number of reasons.
Current loop implies that one or both ends of the current loop are isolated from earthing (grounding to some common entity -- the earth, building steel, the EGC wire, etc.). Thus, you are dealing with a loop of wire for the current loop, a current source, a current modulator, and a current sensor. If this loop is grounded at most at one point, then ground path current can not affect the current in the loop. The shield is not part of the current loop path.
Noise currents can enter the the loop via inductive or capacitive coupling. By using a twisted pair for the current loop wiring this minimizes inductive coupling. The twisting does a phase reversal and thus cancellation of inductive signals covering a large area. Capacitance reactance coupling into the current loop is generally not large compared to the loop impedance. Thus, not likely a problem.
Bad wiring, bad sender (source of current modulation), bad current source, bad current sensor are all likely causes of the problem.
Using a constant load current device, like I described, in place of the present sender and if the sensed results were stable, then this would likely eliminate everything except the sender.
.