Using neutral profile if the photo is taken just after sunset the histogram fills the whole spectrum. Reducing the exposure would mean a significant clipping of the shadows. Is that correct?
Yes, that is correct that if the whole histogram is filled on neutral, that reducing exposure results in shadows clipping. This can be fixed by setting the blacks to a negative value, and dragging the shadow compression slider up.
Pushing the shadows after doing previous step creates some undesirable effects, if you go too far, like halos or these kinds of things
What tool are you using to lift the shadows? What version of Rawtherapee are you using? Versions older than rt 5.0 used to have really terrible Shadows/Highlights tool, but the most recent versions should be a lot better and now can be pushed harder. If you want no local adjacency effects, for a more natural look, use the blacks slider as previously mentioned. If you want a crisper, more contrasty and detailed result whilst capturing the whole dynamic range, use dynamic range compression and Shadows/highlights.
Usage of tone curve 1 and 2. Is there any optimal way, I mean, it is better to change it in 1 in 2 or a mixtue of both. Are you using the auto matched curve? Neutral profile have an straight line. It is better to use neutral, or other bundled profiles
Tone curves 1 and 2 both do the same thing, so the recommended workflow is to use one for smooth rolloff darkening of the shadows, and the other for highlights, but you can get by with just doing an S-curve on using just one. Use auto-matched curve, if you like the tonal response of your in camera jpegs, but don’t if you don’t. Since you are using ETTR, your in camera jpegs will not look the best, but if you are setting the exposure after the ETTR closer to middle grey, auto-matched curves might be a good starting point.
should I activate highlights reconstruction?
Yes, you should always activate highlight reconstruction, and either use blend or color propagation, whichever looks better, which will be image dependent. The one exception would be if you are batch processing lots of images with absolutely no blown highlights whatsoever, to save processing time.
In the tutorials I show I never show that people reduce dynamic range. Can you mention why this would be necessary.
Could you please clarify? Are you referring to using some sort of tonemapping tool to compress the dynamic range, or are you referring to ‘loosing’ dynamic range by adding lots of contrast and causing things to clip black or white?