All of this has been most interesting, and I do thank everyone for taking the time and trouble to contribute to this debate. Now, you are welcome to correct me if I am wrong, but essentially there seem to be the following options if one wishes to use a digital camera or MF digital back and produce not colour images, but B/W prints.
1) Convert to Grayscale from RGB in Image>Mode>Grayscale. This actually is not a straight desaturation, but is more like 30% R, 60% G, and 10% B (according to the info I have found).
2) Desaturate the RGB image in Image>Adjustments>Desaturate. Or desaturate in ACR or other RAW converter using the saturation slider set to zero. This should produce a file which has exactly equal quantities of RGB.
Both of these strategies produce a file which, to my eyes, is rather flat and lifeless. Others may disagree.
3) Use Channel Mixer in PS to mix varying proportions of the R, G and B channels to a monochrome image. This can be understood in two ways - either as attempting to replicate the effect of using coloured filters on the lens as we used to do when shooting film, or attempting to replicate the non-linear response of B/W film stock, or indeed both at once. Most, if not all, of the other plugins, things in PS or CS4 or whatever are just variations or more sophisticated versions of this basic theme. (Some of them are actually quite good - the FP4 profile in CBW Pro is pretty believable. If you like FP4, of course.)
4) Convert the RGB image to Lab Color in Image>Mode>Lab Color and then extract the L channel. CIE Lab seems to produce something genuinely different which is easily differentiated from the other separation methods when you make a print on good-quality stock. Whether you like it better or not is a matter of taste.
Of course, all the above are just the beginning of the process for a fine quality B/W print. Thereafter must follow all the usual adjustments for levels, local contrast, dodging and burning just as we used to do in the darkroom. Essentially, what I am looking for is the perfect negative as my starting point, except that now it is a positive. Following Ansel's practice, it should be correctly exposed, retain the crucial highlight detail, have excellent tonal separation in the all-important mid-range, and have normal contrast. As he so rightly pointed out, we can fairly easily add contrast, but it is much more difficult to reduce it if the crucial upper-mids have been lost.
John