The non-multiband compressor is essentially a single band compressor, what this means is that it acts in the same way on the transients across the whole frequency range. If in peak reduction mode you have some loud transients being played that make the signal exceed the threshold, the frequencies of those transients - let's say it is mostly a bass tone in the piano - then mostly those frequencies are in the transient that is affected by the signal reduction. What this essentially means is that you should imagine the moment of the peak reduction as a series of average looking still images of the frequencies of the sound source at those moments, with a horizontal line (threshold) cutting through it, everything above it is being compressed. So if you for instance want a rounder low mid range, it might not be smart to just feed the signal into a compressor, instead you need to study how the compressor is going to react on the input content, by understanding the input content. Chances are it will actually hit on other frequencies instead, at which point you would be forced to lower the threshold until it starts to hit the low mid frequencies too. At that point you might have already killed the sound source by the compression on the other frequencies in the frequency range.
Due to this you often want to control this further, either by filtering/controlling the frequency response before the single band compressor or by using a multiband compressor instead. So let's say you want a slightly softer and fuller kick drum and you happen to have a kick drum that is quite sticky/hollow. This could yield a pretty crazy compressor configuration until it touches on the frequencies it needs to touch. Therefore you need to prepare the incoming signal somehow first or even better go back to re-recording the right frequencies, so that you can make the compressor hit particularly the low mid frequencies on a low ratio + soft knee type of configuration.
When you use the compressor, it is not the threshold you look at, you look at the signal attenuation meter and listen at the impact of it at the same time. You adjust the threshold until you get the signal attenuation and sound you are going for.
Please also note that you always have the option of splitting up the frequency bands of a sound source into multiple volume faders in parallel with the original track, then target the individual bands with compressors individually on those bands. Be moderate and precise about this due to the added phase inaccuracies that you get from it. Become good at knowing in what frequency range various types of sounds sound great with various types of compressor configurations applied. When mixes have the compression very concentrated on certain frequency bands, it means that those frequencies have been congested to the point that they are now flat. This is why in mastering you work with both input track/bus compressors but also expanders on the mix bus, you want to undo some of that congestion and also just tune the dynamics until it is in balance across the whole frequency range. This is what makes a mix musical sounding, especially when you succeed well with the mid range. This process is controlled by the monitoring process.