I've tried out editing in Lab space white balancing the blue cast out of sunset lit limestone rock while not affecting the blue in other elements in the scene such as flowers. Far more intuitive and smoother than using selective color in RGB.
An overall inconsistancy in hue and saturation levels permiates through different areas of the image giving it an odd look.
I'm having some trouble understanding what's happening here. If limestone is lit by the setting sun, one would not normally expect it to have a blue cast. If anything, the stone should be rather yellow/orange. This raises a question about the quality of colour balance in the scan itself. I've scanned tons of colour negs (in fact written two articles about it on this website) and there is no question in my mind that if you don't get the colour balance nearly right at the scan stage, it is difficult to rescue it in Photoshop, regardless of the colour space used.
Anyhow, returning to the blue in the rocks - you want to neutralize that without also neutralizing the blue in the flowers, but then toward the end of the story you report getting inconsistency in hue and saturation through different areas of the image. That comment raises a question about whether you achieved your objective of using Lab curves to eliminate blue from the rock while preserving blue in the flowers. If you succeeded in doing so without using a layer mask, I'd be curious to know how you shaped the *b* curve to achieve it. One possible approach I could see to this challenge is that if the shades of blue between the flowers and the limestone are different enough, you could, in Layer Styles", isolate the impact of the curve movement to the blue tone on the rock by adjusting the "Blend If" sliders for the *b* curve. Anyhow, however you did it - do tell.
I agree with you that playing with *a* and *b* curves in Lab puts colours on steroids and control is *not easy*, especially when it comes to using them for rebalancing colour. They are much easier to use for saturating colour while retaining the same colour balance, but rebalancing in a way that gives you the "correct" result accross the tonal range for both the *a* and *b* axes is very challenging.
Now, reverting to Jonathan's suggestion made assuming you were adjusting a digital capture, his idea can also be applied to a scanned negative, either by producing two scans with different colour balance settings and layering them, assuming the scanner is dead-on consistent pixelm by pixel from one scan to the next, or by creating a duplicated image layer which you would mask in one way or another, and clip Curves in RGB space to each of the background and the duplicate for the colour balancing, isolated to the effect intended for each layer. I suggest this, because Selective Color, which you mention, is not the first place I would go for altering the colour balance of an image. I would start with the individual R, G, and B Curves.