From what I've been reading a smaller sensor should generally give a shallower depth of field. The main reason being the pixel size of the sensor. Pixel size affects the circle of confusion (COC) which in turn affects the perceived DOF.
Changing magnification changes the focusing distance which in turn changes the relationship of the distance of the lens/image plane to the background which changes how out of focus the background is.
Using this online calculator you can easily see the relationships between COC magnification and focusing distance. Note due to the numeric precision the calculator "breaks" from divide by zero errors of which there is no error trap other than not updating values on the display.
Lens Magnification and Depth of Field Calculator – kielia.de
For instance using a (min focus distance taken from 50mm f/1.7 Pentax-A) at f16:
extension = 0mm
focusing distance = 0.450m
magnification = 0.15×
DOF
COC in mm ) in meters:
FF (0.029) = 0.477-0.427 = 0.050
Crop (0.019) = 0.467-0.434 = 0.033
extension=25mm
focusing distance = 0.210m
magnification = 0.65×
DOF
FF = 0.212-0.208 = 0.004
Crop = 0.211-0.209 = 0.002 (2mm!)
That's why I suggested mounting the camera on a tripod and keeping the subject to sensor constant using the same lens on a FF vs Crop body. The difference in DOF and bokeh would be due mainly to COC and pixel size/density differences.
By changing the magnification as you are doing you are introducing other changes such as subject and background distance to image sensor. By magnification I am using it in the sense as defined for macro photography - the ratio of the object size vs the image size on the sensor plane. At 1× magnification a 10 mm object would have a 10 mm image on the sensor. Thus with the images with the scale or ruler in them at 1× FF would show 24mm top to bottom vs 15.6mm for crop.
Of course all this is leading down the road of "equivalency" of crop vs FF which maybe we don't want to go there. Then again such experiments as these might give us a better understanding.