Drift of thermocouple elements

[Roman oh]

I have finally dusted off my Fluke 540B, and made a little battery eliminator, and have started playing with it. Having satisfied myself that all the circuits are working correctly, I watch it in DC transfer mode (DC from a Fluke 332B which I am confident is far more stable than the effect I am about to describe). The unit nulls correctly, reversal error with the original thermal element is within spec, but over a period of several minutes, the null detector needle drifts to the left - about 5-10 minutes to end of scale in high sensitivity mode (about 0.01% of FS).
If I apply 7mV from a stable source in lieu of the thermocouple, the reading remains rock solid. The reversing switch in the A54 appears to be repeatable and work as expected.
Now, I happen to have a small number of vacuum mounted thermocouple elements and octal bases in my “junk box” (some look unused, till now) so I made up some adapters and tried some of these plugged into the socket of the A54 voltage unit of the 540B. To a greater or lesser extent, they ALL seem to show some drift over a period of minutes - some positive, some negative. As Murphy would have it, the one with (almost) imperceptible drift also has a shocking reversal error.
Is this a common issue with vacuum mounted thermocouple elements? Is my small cache just junk? Or should I be looking elsewhere for something?
Appreciate any help or advice.