Highlight DR doesn’t exist at sensor level with CCDs and CMOS.
Agreed! The is just dynamic range (with a rather ambiguous bottom end, deep in the shadows.) The idea of "highlight DR" and "shadow DR" is simply a split based on how much of the DR is above the midtone placement and how much is below, but of course photographers can move that placement around: see several thousand posts about "ETTR".
At camera level manufacturers can play with ISO and meter calibration (and raw converter smarts) to under-expose to give the illusion of a highlight recovery capability.
A very mysterious use of the word "underexpose": if camera A's light-metering recommends an exposure levels that blows highlights as soon as they are more than three stops above meter midtones whereas following camera B's metering blows highlights less often, some people still say that the camera B is underexposing, whereas I am tempted to suggest that camera A is prone to recommending overexposure! But that would also be wrong; with the variable gain of digital cameras, the idea of one true and correct level of exposure and gain is nonsense. It is in part driven by the DXO "SSat=ISO" error, in which an ISO 12232 standard guideline for
minimum safe exposure index, as in a recommended
maximum amount of exposure (SSat, a.k.a. "base ISO speed", placing metered mid-tones at about 12.5% of FWC, about three stops from the top) is misinterpreted as a mandated
correct exposure level. Perhaps the misunderstanding is enhanced by the myth that the ISO "REI" standard opens the door to arbitrary choices of exposure level, wheres in fact it is simply there to allow for fancy "pattern light metering": in simple center-weighted metering modes, there is clear evidence that the metering on cameras is always close to ISO SOS (Standard Output Sensitivity) — otherwise OOC JPEGs of simple low contrast scenes would look too dark or too light. And that is anyway all about default JPEGs, with the 8-bit JPEG standard forcing an undesirably low amount of headroom between midtones and maximum level; only 2.5 stops.
Given that good modern ILCs having ten or more stops of photographically useful DR (and thirteen or more stops of total DR above the noise floor), and recalling all the agonizing in the early digital era about how digital cameras are vastly inferior to film in highlight handling, it mystifies me that people are still arguing that the one and only one correct way to use that DR is for the light metering system to recommend an exposure level that gives a bare three stops above metered midtone level and seven or more usable stops (ten or more total stops) below. More so at higher EO settings, when additional amplification to raise midtone raw level placement to that "-3" level does little or nothing to improve shadow noise.
And those old CCD MF backs were in fact very close to "ISO independent" as fas as shadow noise, because any variable gain came off-chip and too late, after almost all noise had entered the signal, and their ADC DR was often far better than in-photosite DR.