don't try to record to peak AT 0 dB. You have a
floor to ceiling range (dynamic range) of 144 dB if recording
24 bit audio (or it might only be 130 dB depending on the sound card)
so it's alright for the mix to be "quiet" because you can
gain that 6 dB or less later through normalization/compression/
limiting with out losing much clearity at all.

After you record the mix, although it's quiet, normalize it to
-0.5 dB or so, then run a compressor over it ratio 2:1, fairly
short but not to short attacks say 10-30 ms, and fairly long
but not to long release times, say 50 ms - 100 ms.
and thresh hold as low as you
can go with out it sounding too sparse or "dehydrated",
something that will get you 4 to 8 dB gain reduction,
just something to even out the levels over the entire mix.
Then after that through on a brick wall limiter and limit to
-0.3 dB to 0 dB with no more attenuation of perhaps -3 dB or so,
if any attenuation at all, so as not to destroy to much dynamics of
the mix while still massaging the loudness of the mix.
If you measure the RMS of the mix and the average is -10 dB to
-12 dB RMS, it's plenty loud. These are only ball park figures
man.

Then you always save your 24 bit file (if I can I shoot for
96 kHz sampling rate, but 44 kHz is ok), and then you can
save it to the f'd up 16 bit 44 kHz wave later after applying
dithering, and then from that the even more f'd up mp3 or ogg
compressed file of 320 to 192 kbps, and it won't sound as
"raspy" and "harsh" and "uneven" as the mix cd you put up.

http://en.wikipedia.org/wiki/Audio_system_measurements