First I use AdobeRGB color scheme (at least in the dRebel), RAW format, and always try to keep the image well exposed, avoiding any overexposure. If needed I dialed negative exposure compensation and reshoot, and always check exposures using the histogram.
The correct exposure and wide color gamut of AdobeRGB reduces the risk of strong colors clipping as in the case of strong reds or oranges, and RAW format allows me to correct some errors or wrong WB during conversion. Also, the TIFF generated from RAW has full resolution color information in the three channels (4:4:4) instead of the JPEG reduced red and blue resolution (4:2:2) so less color artifacts are present.
Usually during conversion (if BreezeBrowser is used) I use normal to low contrast, saturation and sharpening (depending of the image), and some positive or negative exposure compensation if needed (in the dRebel RAW files), and save the image as 16bit TIFF.
As you can see I don't use the more saturated colors or increased sharpening used by Parameter 1 (the standard 300D/Rebel setting) but a more controlled color during all the stages of the photo taking and processing.
Lately I've been using C1Rebel with the dRebel RAW files. I also use the ETC 300D ICC profiles instead of the C1 supplied ones.
When C1Rebel is used, I usually use all parameters by default, except noise reduction (I set it to very low), and only made some adjustments to the exp.compensation and contrast sliders, and WB correction if needed, and save it as AdobeRGB 16-bit TIFF.
Once the AdobeRGB TIFF is generated, I load it in Photoshop and then the fun begins
First I check the levels and reduce contrast if needed to keep all three channels in sRGB gamut space. I do this simply seeing in the monitor if there is some color cliping sliding the white point, because the monitor is already sRGB calibrated. When the colors looks good, I convert colors from AdobeRGB to sRGB and readjust levels. Then all selective curves and levels begin, and some highlight/shadow corrections if PS CS is used, and finally the cropping.
Once all is ok, I transform the image to 8bit and apply some sharpening. If the image is already sharp (as usually it is when the EF 28-135 is used) I apply the "clarifier" setting (400-500%, 0.2, 0-2 depending of the image noise), but if the image is soft, I use some more aggresive (200-300%, 0.3-0.5, 0-3 depending of noise again).
And that's all. I save the final image as quality-10 JPEG to keep full color (4:4:4) resolution.
To publish in the web, I took the JPEG, rescale it, apply a weaker variant of the "clarifier" sharpening, using only a 200%, and save as quality 7 to reduce size.
The S50 RAW workflow is similar, except only BB is available, as no AdobeRGB is available, there is no exp.compensation, and all the workflow is done in sRGB. Sharpening used in almost all cases is the "clarifier" setting.