Current Canon RAW files are 14-bit (luminance of each pixel has 2^14 possible values), so a 16-bit file retains all of the data. When you reduce to 8-bit, you are basically throwing away the extra 6-bits...you can't get them back (like setting the mode to grayscale permanently removes the color data).

You won't see a difference on your monitor, since all but the uber-expensive ones (e.g. Eizo) are limited to 8-bits (or less for some cheaper LCD panels). Printing has similar limitations, particularly since most commercial printers want JPGs which are 8-bit.

To the extent possible, I'd say do all the operations you can in 16-bit space before reducing to an 8-bit file.