afaik, the recording level is usually bounded by two things, noise floor and dynamic range.
Noise floor is always there, its just at a lower volume so you don’t normally hear it.
If the native noise floor of a recorder is -80db and the mic noise floor is -60, then your final noise floor is actually the poorest thing in the audio chain; in this case, the mic at -60db.
When do you start hearing hiss then? Some mics need phantom power and a lot of it to get above the noise floor. Other plugin
power mics need a strong signal to get above the noise floor. All of this is very relative. That is why you can’t record something far away and simply turn up the gain.
The other extreme is distortion or voltage levels outside the parameters of either the mic or the recorder. In the digital
world, its a hard limit. analog can push a little past 0db.
So, the best way to answer your other question is that ‘normally’ the human voice has a certain dynamic range when it whispers, talk normally, and yells. -23db RMS is a good place to start because it lets a person talk both softly and loud without having to compress/limit or clip the audio which is probably what happened to your -10 recording). An added bonus is that the normal talking will usually automatically top out around -12/-10db making post a cince as youtube is -16LUFS and theatres are -23LUFS, all your audio will already be close to the target and be the right volume when you sit in the chair.
There’s some sound devices that can analog limit though. So, you may be wondering, “why don’t I just record everything at -1db and let my sound devices soft limit for me?” The answer is, beyond a certain ratio like 4:1, voices will sound unnatural. I don’t go beyond 4:1 sfx and 2:1 dialogue personally but your mileage may vary. People’s ears get tired when dynamic range is too large in a 2 hour movie. Perhaps you’ve noticed this in some independant films.