Interesting to hear of your experiments: I may have to try something similar myself, to see if I can improve Caroline reception.
One thing I've never understood about many of the MW frame-antenna designs seen on the Web is that they always seem to space the turns out over some distance, which must surely make them more susceptible to picking up electrostatic noise and also reduce the sharpness of the null. I'm more used to loop-antennas being wound 'tightly' with the turns on top of each-other, and provided with a Faraday screen so they don't pick up the electrostatic component.
See the DF loops fitted to WWII-era aircraft: they invariably shielded the loop.
http://www.rfcafe.com/references/ele...ary-1963-1.jpg