Brendan, the quality of your recovery looks good to me. I don't recall ever saying that you could not rescue an image by increasing the exposure in post, and your flexibility to do that is increased when (a) you are shooting at low ISOs to begin with; (b) not underexposing too much; (c) have a large enough subject that you are not cropping aggressively in order to get the composition you desire; (d) use a camera that has large, intrinsically clean pixels to begin with.
Quite simply, by underexposing an image and then recovering it later you are having to push up your signal from within the noise in the shadows, but in so doing you are pushing up the noise itself. Think of it like turning up the volume on a quiet tape recording or a quiet section of an LP. The music might get louder, but so does the hiss and the crackles and pops. Given the choice, it is better to record your signal correctly, or a little loud (but not so loud that you get without distortion/clipping) so that if need be you can turn the volume down and reduce the intrusion of the unwanted noise.
In addition to the above, if you refer back to the article on ETTR, you will note that as well as suffering increased noise, you are also going to record less image detail as well, because your important tones will have fewer discrete steps in the RGB values with which to describe your pixels. If you then try to brighten the picture to compensate you will be stretching out the tonal scale, causing gaps in the histogram, which if pursued too far can appear as posterisation in the image. This is less of a problem with raw files than JPEG, and may not necessarily be visible, but the issue is a potential risk.
If you have a scene with an extreme dynamic range then you may have no choice but to end up with the shadow details underexposed, simply to protect the highlights. In such a situation, unless you can reduce the dynamic range, perhaps with filters, or by adding light to the shadows, you will have to do the best you can, and perhaps compromise your highlights a little if shadow detail is more important. However, in a scene of low dynamic range, there really is no reason to be underexposing anything, and best practice would be to avoid it.
Here is an example from my 7D at 1600 ISO. The attachments are....
1. Full image, no edits. The noise in the file is not really offensive because I have reduced the file to approx 20%. If the subject had been small in the frame, requiring a heavy crop, the story might be different;;
2. 100% crop, no edits. The noise is pretty evident when viewed at this scale;
3. 100% crop, exposure increased by +1. If we pretend that the original was underexposed and we did actually need to increase exposure by 1 stop then this would be looking pretty ugly right now. You can imagine what Alamy might have to say about such a file;
4. 100% crop, exposure reduced to -1. If the original file had been purposely overexposed by a stop, while still preserving highlights, this is how the noise would diminish simply by correcting the overexposure. There is quite a difference in outcome between underexposing by 1 stop vs overexposing by 1 stop. Being able to overexpose (ETTR) and then to reduce the exposure in post is a pretty powerful tool in the war against noise. Also, given that no NR is required, or reduction in sharpening, the details that are there are fully preserved rather than being massaged away to obscurity.
