The idea of having a few saturated pixels along with a few background pixels at zero does one thing - it maximizes image contrast. Image contrast has nothing to do with quantification. This is a distinction that is very important and also very misunderstood.
When you are adjusting parameters on your confocal microscope - you need to keep in mind that you are optimizing raw data for potential image analysis - so ideally you want no overexposure or underexposure anywhere in the final image (no green or blue pixels for Leica, no red/blue for Zeiss, etc.) for all of your conditions. This way you can ensure that all your measurable signal falls within the detectable range of the PMT.
Often this will not get you the best image contrast - but at this point you should not be worrying about this. Image contrast can be adjusted in post for optimal display.
It is not impossible to adjust your system so that you have no overexposure in your images for every condition - you just have to have a good idea of what to expect. This is where controls come in. Positive control conditions give you a good idea of how bright your signal should be - and if you make your adjustments to your brightest condition - the rest of your conditions should fall somewhere below the maximum. Or if you do not have a positive control - you can underexpose your brightest condition by 10-15% to hedge.
Sometimes this is a difficult point to get across as you may have a hard time visualizing good image contrast in scenarios where you have a large signal disparity between your brightest condition and your dimmest condition. That does not mean that the disparity is not quantifiable - just that your monitor may not be able to display all the data contained within the image at one time. Don’t let that trick you into overexposing however. Most monitors cannot display the full dynamic range from a 16 bit image.
The main point I make with my users - if you overexpose - you cannot get this data back. It is gone forever. If your adjustments collect all the data - it can be further processed to render a measurable data set AND a high contrast image from the same data.
Hope this helps!