Different scanner software behave differently.
After a preview you always have the "mountain graph" (the histogram graph). With certain scanner software (such as ScanView) you can have the basis of the mountain to coincide with the extremity of the graph field. So the darkest zone on the slide (let's imagine a black and white slide) will be rendered as 0 and the brightest will be rendered as 255.
That regardless of whether the darker zone is actually pure black and the brightest is actually pure white.
With some other software (NikonScan e.g.) you see the "mountain graph" but the mountain slopes do not necessarily reach the extremities of the field. This means that the darkest zone is not "0" and the brightest is not "255".
If you manually move the black point cursor rightbound, and the white point cursor leftbound, until they are at the beginning of the histogram, you are recreating a situation where the darkest point is 0 and the brightest point is 255.
If you do so in such a situation, you are obviously increasing the contrast of the scan.
If you do or don't do that is something which is left to your aesthetic appreciation, but you have a technical reason to do that anyway.
What you have read or dreamed is probably this: when the slopes of the histogram do not really arrive at the extremities of the graphs, if you adjust the white and black point cursors so as to put them just under the "mountain" extremities, you will as a result use the entire "numeric space width" of the scanner (be it 8-bit, 12-bit, 14-bit, 16-bit) on useful information from the slide.
If you don't do that, you are "wasting" part of the bit-depth of your scanner because the entire image info will be constraint into a "subset" of the numerical space that the scanner is able to represent.
To put it simply, let's say your scanner is 8-bit and your mountain is very narrow. Without any adjustment you would obtain from your slides let's say values from 15 to 225, so you will only have 210 shades of grey. If you adjust you obtain values from 0 to 255, higher contrast, and you will have 256 shades of grey. You will certainly have less problems in postprocessing (less posterization, less quantization error on resizing, better noise reduction) with 256 shades of grey than with 210.
As a final stage, you can always reduce again contrast in Photoshop if you want to obtain a less contrasted image.
So the logic is: use as much as you can the bit-depth of your scanner, by moving the black and white cursors (and increasing contrast) at scanning stage, so as to have a scan which suffers less during postprocessing, and if necessary reduce contrast again at the end of postprocessing.
If you like a comparison, that is a trick equivalent to the "expose to the right" with digital captures: you are adjusting your capture in such a way that you have more information to work with during postprocessing. You don't care about the appearance of the capture being "wrong" because you are doing postprocessing anyway and you will readjust things.
The entire problem is much less problematic with modern 14-bit or 16-bit scanners which do not really suffer that much of quantization errors (they do the same, but the errors are much less visible, if at all).