[TonyB]

You're right, those peaks can limit the overall output (apparent loudness).

Mastering is a process that typically consists of some subtle (hopefully subtle) EQ, Compression, De-Essing, and Limiting.

Basically what I do after EQing (if needed, maybe some high shelfing), and removing unwanted noise (crackles/pops) is to add some compression (not like what is done on individual tracks during mixing) to tame the peaks and bring the mix together as a whole.

Typically what I do is:

- EQ: I typically use the UAD Cambridge EQ; not much but usually focus on the low and mid- freqs at this point

- Remove Noise: Remove crackles and pops ( hopefully I got these during the mixing, but sometimes those nasty noise show up during mastering.

-De-ess: I don't always use this; depends on the recording and mix

- EQ: I'll go back and put some high shelf EQ

- De-ess: Check again to see if high shelf EQ added any sibilance to the vocals or brought out anything else in an "unnatural way"

- NOW TO YOUR QUESTION: This is where I tame the peaks. Two options:
-- Manually
-- Automatically

-- Manually: Sometimes I go through and manually bring down each peak one at a time volume automation. You need to really zoom in on the peak. This is actually compression, but you're manually doing it. I'll use the Volume Envelope to bring the peak down. I'll pull down the peak to somewhere above the overall (perceived RMS) mix.

-- Automatically: I'll use a plugin to automatically "bring down some of the peaks." Usually between 1.0 and 2.0 dB of leveling with a very slow attack and release. If the attack is too fast you'll quickly lose dynamics and punch to the song

- Limit: Now I'll put on the limiter (always last. It depends on the song and genre, but for alot of what I do I use limit at around 1.5 dB at -0.1 dBFS

You're right, bringing down some of those peaks will bring up the overall volume and fullness (hopefully) of the mix