The realization that slapped me in the face is that there really is no “correct” exposure. For a low-dynamic-range scene, you can “move” the raw data across the face of the sensor a few stops for a set of images, and when in post you anchor each’s highest value to white, they’ll all look pretty much the same.
It’s the middle gray reference to which most meters are anchored that determines how the rest of the data lays on the sensor. What’s the number I saw recently, about 2.4 stops from gray to saturation in most sensors, so any light that goes past that will just saturate the sensel. In Watership Down terms, the “hrair limit”… 
A lot of the time, the data past the hrair limit is just specular, in-scene light sources. But sometimes, it’s brightly-lit objects; go see my “Afternoon Snack” PlayRaw for a good example. In those cases, one might want to underexpose from the camera’s assertion, pull those bright bits into resolution. But that will push shadows down toward black, and into the noise floor of the sensor, if they weren’t already there. Sometimes, the photographer is hungry, and wants to eat more than wanting to mess with the exposure…
Soooo… pick your poison, I guess. This is why I’ve bought each successive camera, chasing dynamic range, so when I pushed shadows further into the well, the sensor tolerated it a bit better…