Thank you for the responses.
The DUT's input impedance appears to be around 2Kohm at 10Khz. So not that high. Although the signal is fairly low, so I wouldn't have thought it'd matter much. We're slightly adjusting the signal generator's "1Vrms" output so that the (very accurate) voltmeter is perfectly at 1Vrms each time.
The DUT is just being calibrated, so there isn't really a DUT output to monitor.
This isn't my design so, unfortunately, no schematic.
If we replace the voltmeter's T connection for probes, in my setup, we get the same result. There is an additional error doing this with the customer's setup.
Correct me if I'm wrong, but doesn't the customer's setup have multiple signal paths? One from the signal generator to the 50ohm terminator, then another from that point to the 2k DUT load. If the voltmeter is placed after the DUT, then there is a third path with essentially no terminator.