It’s not that simple… Loudness is realted to the average level and some other factors, whereas clipping is determined by the peak level. Many quiet-sounding songs have one or more 0dB (100%) peaks, so you can’t increase the loudness without clipping (distorting). “Equal loudness” and “loud” are often conflicting goals!
which I’ve used MP3Gain to ‘apply max no clip gain for each file’. The trouble is they have wildly different volumes now.
If you don’t allow clipping (and you usually don’t want clipping) there is obviously a limit to how much MP3gain can boost the volume, if it can boost it at all.
If I make the first file 88.4 dB to match the second file they are both too quiet.
Since we have a 0dB peak limit, most of the time if you want to match volumes, you have to reduce the loud-song song rather than boost the quiet-sounding song. That’s just math… It’s not MP3Gain’s “fault”. MP3gain (and RelayGain, etc) will end-up reducing the volume of most songs. You can increase the target-volume for MP3gain, but you are giving it less “room to work”, and you’ll end-up with a bunch of unchanged files (if you don’t allow clipping).
I see a big peak in the “quiet” file. You can try some dynamic compression (Effect → Compresssor) or a limiter plug-in to knock-down the peak. Compression/limiting is most-often used to boost the average level without boosting/distorting the peaks. If it’s over-done (as it often is with popular music) it can make music boring by making it “constantly loud”. And, if you really over-do it, you can get distortion or other side-effects.
There’s one more thing you can do… After running MP3gain, run Effect → Amplify. Note the default amplification and cancel. The default is the amount of gain you can add (if any) without clipping. Choose the smallest default amplification of both files, go back and re-run Effect → Amplify, applying that smaller gain to both files.
It will also help you have original uncompressed files, so you can apply ReplayGain (or WAVEgain) before MP3 compression. MP3gain works only in 1.5dB steps (due to a limitation of MP3), and the lossy MP3 compression causes some peaks to be higher and other peaks to be lower. The higher peaks effectively force a lower gain-before-clipping limit.
And, any editing in Audacity (or any “regular” audio editor") requires the MP3 to be de-compressed. You go through a 2nd lossy compression step if you re-save in MP3 (or other lossy format).