The jpgs coming out of your camera are far from raw. The camera makes as many choices as a retro camera filter does, if not more. It chooses color balance, color space, contrast/response curves. It reduces noise, sharpens, and can even adjust the final exposure. There's no one way to display a RAW file on an 8 bit per channel monitor. Furthermore, the "accuracy" of your image is going to depend on the situation and whether or not you override the metering. For example, a snowy scene will almost always be metered into a dull grey, unless the camera is told to overexpose a bit.
My camera is "prosumer"; I have more control than a point&click, though it's not "expert" class. What gets cut off isn't important to me. There is no unbiased data source, it is literally mathematically impossible, so that's not the standard I use. Going back to audio, just because I like digital fidelity does not mean I'm obligated to be snooty about getting the full 24-bit/192KHz digital masters or it's just not worth it; 16-bit is good enough for most use cases I have, few of which involve listening to music in utterly silent acoustic environments with my yearly salary's worth of reproduction hardware.
You're missing what I'm saying. I wasn't speaking about fidelity at all, I was trying to explain that conversion from sensor data to JPG is a matter of stylistic interpretation, even if it is done automatically.
For example, if you have a Canon, you likely have picture modes like Neutral, Landscape, Portrait, and Faithful. Or, if you have a Panasonic, you'll have modes like Standard, Dynamic, Nature, Vibrant. While you may consider one of these modes to be more realistic or take Canon's word at their concept of 'neutral' you'll find that, for some cases vibrant or nature is actually more accurate.
Even a monochrome color mode is likely to be more complicated than just equally averaging the red green and blue values. It will probably have stronger response in the reds, and the least response from blues because of the pleasant effect that has on skin.
There's nothing wrong with letting the camera make those choices for you, but the jpgs your camera spits out are tweaked in ways that are intended to be pleasing, not necessarily the most accurate.
To me, this is quite similar to choosing a particular film or a post-process filter that you like.
I'm not sure you got my point. There is no non-stylistic interpretation, though I used the more technical term "biased" in the machine learning sense of the term. There isn't even one mathematically. Even RAW is inherently biased by the nature of the CCDs. Even what you physically see is fundamentally biased by the nature of the human visual system, which makes an amazing variety of choices for you long before it reaches your conscious mind. I don't sweat the fact that I'm not getting something that absolutely can not exist.
Your point is meaningless, because there is no way to choose a system such that it isn't affected by your point.
