we have hit the brick wall.

One of the confusions about DR is equivocation - the same word being used to mean different things. There are two main definitions of DR common in photography. One is the one you're giving, which is more properly 'tonal range'. The other is to do with information theory - what is the range of information available in the raw file. You can't really talk about 'clipped blacks' or 'clipped whites' in a raw file because until it's processed the black and white levels have not been set. The pixel saturation gives the highest possible white level, but rarely is white set there in a processed file for various reasons. At the low end the decision on where to set the black level is usually about noise, not hard clipping.

Not my camera, one I'd been sent the raw files for from AP to do the analysis. On the 17 stops, first remember that was an old camera - before the dual conversion gain was being used. Dual CV usually reduces the maximum DR because it increases the read noise at low ISOs. Its use really is to extend the available DR at higher ISO settings. Also, read noise tends to increase as ADC speed increases, and with cameras going for really fast readout it has increased substantially. That camera (an A7II) had a base ISO read noise less than a quarter of the A7IV
On the DR figure itself, it's not remarkable - remember that many cinema cameras claim that kind of DR - it's not that their sensors have any special juice - and the actual DR figure depends on normalisation, which is vital if you want to compare cameras. The likely reason that my figure is higher than Bill's is the way the read noise is measured. The DR is very dependent on that, because it's the denominator of the fraction. Mine is obtained by measuring a black frame (taken with the lens cap on), whilst Bill's is obtained by curve fitting and extrapolation. It has to be done in that way for him because of the way he gets the data, but a direct reading will clearly be more accurate.

"must have"...
Anyhow, it all depends just what it is you tested and how you tested it. What does 'clipped both ends' mean? Were you analysing the raw file or working with a raw processor, if so which one? How did you set the black and white levels? And so on.
The figures I gave don't say what is usable, they just show the extent of information content. What you consider usable and how you use it is up to you.

I'd never use a DRO setting if working with raw. DRO is about OOC JPEGs really. Generally they set up the sensor to reduce read noise (I'd guess that Sony would switch in the higher conversion gain) and expose brighter than they would normally for that setting. The high conversion gain would track with what you found.

ACR's doing a lot there. To really find out what the camera's doing in terms of potential tonal range you need to do a linear raw conversion (for instance, with dcraw or one of its successors) and two different lightness adjustments to find out where at the top and bottom ends you can differentiate tones. You can't do it with one because the camera will almost always give more tonal range than the viewing device you're using.

Exactly right. The noise dithers. Just like your printer can only print black and white (if you've got a cheap one) but can still produce a range of greys.