There is no any standard amount of a protein to do a relative quantitation by densitometric analysis after SDS-PAGE. I would use a low amount of your protein to make sure that other bands almost disappear (in ideal case the other bands reading should be close to the reading of the background, where there is no protein bands). The general rule is that you need to reach conditions where your readings will give you the highest ratio of signal (your band) to noise (background). And always subtract the background reading from your band reading.

I agree that there is no standard amount for quantification purposes on SDS-PAGE. However, our Quality Control department works closely with the FDA for our vaccines that are in clinical trial, and we are required to report the purity of our proteins by densitometry having 2 ug of protein per gel lane. So I would highly recommended trying that.

It also depends on the molecular weight of your target protein and the glycosylation extent (if glycosylated). As a general rule, the amount of a protein with 10K molecular weight should be loaded with 50-fold less amount than a protein with 500K molecular weight (in micrograms) to get an equal signal. 

If you really want to really quantify the % purity of your protein using SDS-PAGE, you should load a range of protein quantities (1, 2, 3, 4, 5 µg) on the gel and scan the stained gel using a densitometer. At low loads, you will underestimate the amount of contamination, and at very high loads you may exceed the densitometer's linear range and underestimate the purity, but you should see an intermediate range of loads at which you get about the same purity. You should use Coomassie staining or fluorescence staining, but not silver staining. Note that if there are impurities that run at the same place as the main protein, they will be missed, causing the purity to be overestimated.

Thank you all for your useful comments. I will follow them and get back if I still get some confusion.