The buzzing sound worries me a little. Before going any further I think we need to identify and eradicate the buzzing sound. Doesn’t matter if you normalise to any level if that buzzing sound is present, even quietly, it will get sucked up in the processing that occurs in the presets.
Before coming on to your workflow what’s your audio hardware? USB Mic or XLR mic? And how are you routing the signal from mic to PC? Are you using a pre-amp/processor or just an interface? Is there anything in your recording area that could be making the buzz? Are you using a desktop stand? If so there could be vibration coming through the desk and into the stand?
It may be possible to eradicate the buzz in post processing by using the scientific filter and rolling off below 80hz. Generally there’s no audio in the human voice that gets that low so cutting off below 80hz can elimimate some bass rumble if that’s what it is but the best option is to try and find the source and cut it out from interfering with the audio recording.
In answer to your question you can record in to Audition anywhere between -20db and -3db and you have usable audio. From what I’ve seen Mike’s preset template has a very good set up that will level pretty much all audio.
I have my levels set to record at -6db. This gives me headroom where I know am not likely to be clipping. My dbx286s has a clip light and that feeds into a Scarlet 2i2 which also has the red ring around the gain control so there’s plenty of warning when setting up your audio hardware if you’re likely to clip.
When setting your input levels firstly speak normally go get a rough idea of where it should be. Then read something in the style of your finished read and again watch for peaks to make sure nothing is hitting the red. Within reason it’s best to record too quiet as opposed to to loud, you can always normalise up the audio but if it’s clipped it’s wasted!
I would suggest recording into Adobe Audition once you’ve identified and removed the source of the “buzz” and put that recording direct into the preset template. It may well be that you don’t need to do anything other than record then process through the presets.
As with anything audio related, poor quality in will hamper badly final quality out. So you want to get rid of the buzz at source. Audition does have noise reduction that could deal with it but see that as a last resort if you absolutely cannot find and erase the source. Noise reduction is world class in audition but it works by identifying (or being taught via capture noise rprint) unwanted frequencies. It then dials back those frequencies which lessens (or removes) the noise but any corresponding frequncies in your vocals will also be dialed back as part of the process. Noise gating or downward expanding isn’t an option here because while it will remove the buzz while you’re not talking the buzz will instantly become audible again when you start talking.
In answer to your original question, if you do normalise then do it to -3db as this leaves headroom for any processing or EQ that you do without the risk of clipping. If you’re using 32 bit for your recording then you can recover digital clipping (passing 0db in digital post production) but the best idea is to leave some headroom to work with.