Had a bit more of a play with the cloud in my tree animation test and noticed that the clouds did exhibit some of this lighting effect. I ended up inthe same area as Matt suggested (and Volker). I think a detailed explanantion of what each of the cloud settings is actually intended for would help, but as has been suggested by others it's also a case of working out what you're actually trying to simulate. This is a lighting effect (so the lighting/tweaks sections will be a start) which is partly dependent on the cloud type (all of the other settings)
One of the main problems I find when experimenting is that when you make one adjustment to improve one thing, you end up making something else worse. To get around this, I usually try to look for relationships between two variables and use this as a base for making large tweaks. It's not an entirely scientific approach but it can provide some decent starting points.
In the case of clouds I'm currently making some progress with these two:
Tweaks: Fake Dark power = Lighting: Fake Internal Scattering
Lighting: Light Propagation = 2x Lighting: Light Propagation Mix
A lot, however depends on the settings Main Cloud settings, the cloud's Fractal Scale and Density settings. That's 3 sets of 3 settings which can make the job relatively complex. One thing that has confused me regarding people reporting how many cloud samples they are using is that the number of samples for a given quality level will vary according to the values of these settings.
This makes it very difficult to actually determine whether someone is using a very high quality setting or not and it would seem to me to make more sense to actually report the quality level. If you're using a quality level of 1 and it's stating an equivalent number of samples over 120 then there's a good chance the image will be noisy and require substantial additional tweaks to the atmosphere and rendering setttings to fix. Most of my clouds only have 60 - 90 samples for a quality of 1 and I never have much trouble with noise.
I'm not saying that there is an optimum number of samples that we should aim for, as this will probably vary depending on the cloud type... e.g. requiring higher samples with thicker/denser cloud settings... but if you find that the number of samples is really high when you set the quality to 1 you may well find that you can make some small adjustments to the 3 x 3 settings mentioned above to lower the amount of noise in the image without affecting the cloud shape too much.
Here's a rough test I did on the train this morning. GI was off for speed, but I'd recommend having it on some for some really cool lighting...