Alcatraz, the key thing about JPG files is that they are scientifically designed to conserve space by compression. In JPGs, compression is accomplished basically in two ways.
1) a lossless type of compression, common to many types of files, where multiple repetitive elements are replaced by much smaller 'placeholder' tokens. So a whole region of identical blue sky, could be replaced by a much smaller array of pixels of tiny code elements. This is how your typical 'zip' sort of compression program works.
2) a lossy type of compression, where information is actually discarded. The JPG standard was developed over a number of years, with research into the psychology and physiology of vision, among other things, and the information discarded, is by design, intended to be unnoticeable to the human eye. Of course, this is all arguable, but that's the intent.
The compression vs. quality curve is not determinate, and can be implementation dependent. There is no formal definition of exactly how the software is to be written, so different products have differing results.
For most purposes, it's trial and error. Fiddle around until you find settings that meet your size and quality needs.
Always remember, despite 1) the lossless component above, there is always some loss of information in a JPG save, so always protect your original, and never work from a JPG saved from a JPG, (perhaps saved from a JPG....) if you can help it.
Additionally you can specify the pixel dimensions of your save. It's not necessary to post a 25 megapixel image on a web site. Even on a high-end HDTV, that's a tremendous overkill of information. Consider the needs of your output medium.
