Different people can perceive on-screen images differently and so will find that different luminance and white point settings provide the best visual match to a proof under correct lighting conditions. This is why it's worth investing in calibration software that offers control over these parameters, and expecting to perform a few iterations to find your ideal settings.
Luminance: If you have more than one display and want them all to match then you'll probably need to set them to the same luminance. Professional quality calibrators will allow you to do this, sometimes with the option to measure the luminance and possibly also the white point of one monitor so that others can be set to match it. The value you set will depend not only on what your displays are capable of but what the ambient lighting is like; although relatively dim and neutral-colored surroundings are strongly recommended, sometimes it's not possible to achieve this. If you're regularly comparing screen with print, it's a good idea to set a visually similar luminance for the screen and viewing booth in order to minimise the adjustment your eyes have to make. Since cheaper viewing booths don't tend to have dimmers, that may mean adjusting the screen. If you're calibrating a mixed group of displays, you'll be limited to what the dimmest can manage.
If you calibrate and profile with software that reports the achieved values at the end of the process and find that the target luminance hasn't been achieved, first try raising the hardware luminance level if possible (i.e. turn up the screen's brightness control), as a poor calibration can result if the target luminance is not reached. If a repeat calibration still does not achieve the target value, set the target lower and run the process again until it can be achieved. Your manual hardware brightness setting needs to allow for some inevitable loss of luminance as a result of automated colour balance adjustment later in the calibration process, so allow some 15 - 20 percent extra.
At the other end of the luminance range, you may find it useful to be able to specify a minimum (black) level, if you want to soft-proof for commercial print, for example. First try the default, which is probably minimum black level and compare it to a proof under correct lighting conditions. If this isn't a good representation of printed black and your software allows you to specify a value, try increasing your targeted black point above zero in small increments, then compare this against a proof. Be prepared for some trial-and-error adjustment.
White point: The D50 (5000K) illumination standard used in the printing world might at first seem the obvious target for display calibration but experience has shown that it can look too yellow, especially at lower luminance settings. D65 (6500K), although technically and visually bluer, tends to look more 'right' to photographers when comparing against accurate and correctly-illuminated prints. If you do find 6500K too blueish and your calibration software allows it, you can always experiment with in-between values. With LCD displays it's best to ignore any colour temperature presets that are offered via the monitor's buttons and on-screen control menus, as they can be quite inaccurate.
Gamma: Gamma settings can be a source of confusion, but in a correctly colour-managed system the actual value doesn't matter too much as it will automatically be taken into account. Since the native gamma of almost all displays is 2.2, from experience we recommend standardising on gamma 2.2.
Verification: A good quality display calibrator/profiler will offer some means of assessing the quality of the profiles it makes. It's a good idea to run this immediately after making a profile but the ultimate test is via visual verification. This is done by comparing a certified proof against the screen display of the same file. The CMnet/Pixl verification kit supplied by colour management net includes a measured and certified image file with the appropriate Photoshop proofing setup instructions. The print needs to be viewed under the correct lighting conditions, ideally a D50 viewing booth, though daylight is usually acceptable - a GATF/RHEM indicator is attached for guidance as to whether viewing conditions are acceptable.
Re-profiling: LCD displays are a great deal more consistent than CRTs, but LCD panel behaviour does change over time, so you will need to re-profile at intervals. Try re-profiling LCD displays every two weeks to begin with. If there's no apparent change after re-profiling, extend this interval to three or four weeks. Even if your display seems very stable, it's good practice to re-profile monthly as LCD panel backlight tubes do change over time. The LED backlights used in newer displays are claimed to be more stable in the long term but this is as yet unproven. If you have one of these displays and find that there really isn't any change from one month to the next, then you could try doubling the testing/re-profiling period, but spending a few minutes each month to make sure that your display colour remains accurate doesn't seem too high a price.
This article is based on Practical Colour Management for Photographers and Digital Image Makers, an e-book that's full of detailed practical tips on how to 'do' colour management on real computer displays to support high quality image editing. It includes tips on choosing and using colour-critical equipment and the optimum system software settings and working environment for reliable results.