Spectral frequency display in adobe audition shows my voice peaks over 55khz

There are several things to say about this. Let's deal with the level issue first:

When you record your voice (or anything else, come to that) you don't record up to 0dB, because what you have above is exactly what you get if you do - clipping. When the signal gets to 0dB the waveform tries to take an abrupt right turn, which it can't do - what this looks like is a bit of a square wave - and those contain harmonics that in theory extend to infinity. And those are the lines you can see in the spectral display extending up to ridiculous values. The norm for recording on digital equipment is to have the highest peaks at somewhere between -12 and -15dB - just to allow for any peaks you weren't expecting not to clip. This is called 'headroom', and it's what all pro engineers do when they record. If you don't do this then you've overloaded the system, and the distortion is inevitable.

The correct procedure is to make your initial recording so that the peaks are around -12dB, and then do a process called 'Normalizing', which can accurately determine where the highest peak in your recording is, and set that point to about -1dB. This is available as a function in Audition. Now it may well be that you need to do other processes before this, like compression and limiting to optimise your voice levels and if you do this, you may find that you need to normalize more than once, but the principle is the same - you don't have any part of your recording clipping at any stage.

It is also possible to set levels inappropriately, especially if you are using a mixer. If, for instance, you set the output fader too low and try to compensate by turning up the gain control, then you get early-stage clipping, but the levels seem correct. Getting this right is called 'gain staging' and it's important to do this in the correct order. To be fair, this isn't explained properly in the mixer's guide book, and there isn't a simple way to set this accurately without a tone generator - which this mixer doesn't have. The chances are though that if you leave the fader set to 0dB, and adjust the gain control so that the mixer's meters read -12dB on peaks, that you'll get the same peak levels in Audition, and that's actually what you want - not the mixer's meters hitting 0dB. And this is almost certainly why you still appear to have clipping, even though the signal is below 0dB - it's early stage overload. With a tone generator it's much easier to set the levels correctly - you just feed a tone level into one of the channels and set the master fader so that the output meters read 0dB. And then check that Audition also reads 0dB when you record. You leave the fader at that point (it should actually be 0dB, but there's no guarantee) and then use the gain control to set the recording level for the aforementioned -12dB peak level. You don't worry about the meters only registering on the bottom two steps - that's correct!

You really don't need to record at 96k, especially with an SM58. All that recording with higher sample rates achieves is to record spurious noise and signals above the range of human hearing, and make your interface and everything else in the system run at twice the speed it needs to. If you are recording for video, the norm is to use 48k, and that will give you a recorded frequency range that exceeds anything that even a four-year-old can hear, never mind an adult.

If you record correctly at 48k (without any clipping) you will find all of that space above about 10k is empty. You've recorded it, and there's nothing there. So why? Nobody can hear it, and it makes all your files twice the size they need to be. There used to be an argument about getting a 'cleaner' high frequency sound if you recorded at a higher sample rate, but that all went out of the window when anti-aliasing filters were dropped and over-sampling started being used. That's now universal, and has been for decades. Upshot? There's absolutely no reason for recording at that sample rate - there are absolutely no other benefits. There are a lot of urban myths about this, all of which have been debunked. There's a good video about this here. 

In short, it makes more sense to record at a greater bit depth than a greater sample rate.