ArtOf brings up an extremely important point when it comes to masking: One often gets much better results, and it almost always is faster to combine two masking tools than to try to milk the absolute ultimate performance out of one tool.
For example, ArtOf suggested starting with a global tool (color range), and then improving that selection using "Refine Edges".
Another very helpful approach is to intersect (Boolean AND) a previous selection with a new, very roughly drawn selection. To give a specific example, suppose you need to extract one object from a scene, but there are two similarly colored objects in the scene. Rather than spending time endlessly tweaking what you might get from the color range tool, use it to select both objects and then run a quick lasso around the one you want with the selection mode set to "intersect".
I also do the same thing using external masking programs such as onOne's ReMask or Vertus' FluidMask to work on areas where they work well, and then use, say, the pen tool or the Polygonal Lasso to quickly knock out sections with well defined, more or less straight edges.
Also, don't forget that if you have a couple of selections in the form of masks / alpha channels, using "lighten" or "screen" to combine them is the equivalent of "Add to Mask", and "multiply" is the equivalent of intersecting the corresponding selections, etc.
One can't be married to any one particular technique - your life will be much easier if you use the best tool for each part of the job and know how to quickly combine the results.
HTH,
Tom