oscroft
Veteran
I've tried both. I reckon that if your target is the the same bit depth (say 16 bits), then scanning at 16 bit gray, or at 16-bits per color RGB and converting to gray afterwards, should get you the same results either way - as every pixel you scan is gray, you should get the same level in each color channel (if it didn't give you that, the gray scan would actually be colored). And when I compare scans done both ways I can't really see any difference.
But the difference comes if you might want to adjust the color channels before converting to gray. I'm not really sure what the benefit of that is (but I know some people do it). But as I haven't had any desire to do that, I currently scan grayscale, with the plan that if I should ever want to try adjusting the color channels in any scans I've done (assuming I find out what the purpose is), I'll go back and make an RGB scan.
But the difference comes if you might want to adjust the color channels before converting to gray. I'm not really sure what the benefit of that is (but I know some people do it). But as I haven't had any desire to do that, I currently scan grayscale, with the plan that if I should ever want to try adjusting the color channels in any scans I've done (assuming I find out what the purpose is), I'll go back and make an RGB scan.