I guess that the faster readout of 9.4ms is achieved by the line skipping method? The Japanese test should not be of a concern, because the difference between mechanical and electronic is obvious. You can just refer to the model name.
FWIW, the shutter speed was set to 1/8000 or 1/4000 depending on the shortest one offered by each model.
No line skipping is involved in this case (although it is common for high frame rate video such as at 100, 120, 200, or 240 fps); the Z6 III video at full sensor readout and full 6K resolution at 24-60 fps has a 9.4 ms readout speed. I think there are two reasons for the slower readout for stills: there are more lines to be read (2:3 instead of 9:16 aspect ratio, the latter being used for video), this explains why the JPG read time is 11.03 ms, the other reason is 14-bit readout takes more time (to get less noise in the least significant bits), this can explain the difference between JPG (11.03 ms) and 14-bit NEF (14.41 ms).
So basically, in the Z6III, stills take a bit more time to read from the sensor than video because the image is taller and the implementation of 14-bit readout to get lower noise and better dynamic range requires more time. In the Panasonic S1II, also using a partially stacked 24 MP sensor, they offer a high dynamic range mode ("DR Boost") with an even slower read time of 27.5 ms. This illustrates the compromise between rolling shutter and dynamic range in video. In stills, a mechanical shutter can be used to mitigate slow readout speeds of high DR sensors, and indeed, the Z8 and Z9 do not quite match the D850's dynamic range at base ISO.
