Hard to kill because it is NOT a myth.
Yeah...it is. Smaller pixels have lower well capacities and lower read noise, implicitly, so there's no implicit reason for small pixels to have lower DR than large pixels. That's why there's no rhyme or reason as to the DR among the range of pixel sizes, from 2 microns to over 8 microns.
A77, 3.89 micron pixels, 12.2 stops
5DII, 6.39 micron pixels, 11.2 stops
5D, 8.04 micron pixels, 10.8 stops
How can you say " there's no rhyme or reason" to something that is a measurable and quantifiable value. There are also reasons, that are in fact very well understood, for the sources and amount of noise in any electronic circuit.
The three cameras you listed are all from very different generations, many years apart in fact. That's a reason, technology improves over time.
Btw, where you write "implicitly" in your comment I will assume you meant to say "inherently". I will also grant you that a lower noise floor allows for higher DR for a given maximum signal and that is because DR is the ratio of the largest signal to the lowest signal that can be recorded. So yes, in a case where the noise floor reduction is proportionally larger than a decrease in the max signal (due to the smaller photosites) then DR would go up.
You might also read my comments again and notice that I identified the Sony A77 sensor as having achieved a notable increase in the full well capacity which is done by increasing the size of the photosite which is the surface area of that is sensitive to the photons. The size values you are quoting are the pixel pitches which is not the same thing... there are gaps between the photosites due primarily to the electronics that sit beside each photosite. It would seem Sony has found a way to increase the size of the photosite possibly by reducing the size of the other electronics... with more area available for the photosite they can be made larger so that each can collect more photons.
Historically, larger pixel sensors have had higher read noise than contemporary smaller pixel sensors. Read noise is actually made up of a number of noise sources in the electronics but not all of them scale with pixel size. It has also been historically the case that the read noise at low ISO was higher than at high ISO and this was related to the noise sources that scale with the fullness of the wells (at high ISO settings the wells are not filling up) and the way the amplifiers work. CMOS sensor technology has advanced read noise has come down which improves DR. Again, the Sony sensor in the A77 is notable because the read noise has come down at all ISO levels and in fact it's a significant improvement at low ISO such that the read noise is remarkably consistent from low to high ISO. If this can be implemented for a FF sensor with larger pixels (higher maximum signal) we should see a big bump in DR!!!
I'm happy to discuss further if you or anyone has any facts or examples about this but please don't just write back "yeah... it is".