What these files show is a sine wave frequency sweep from the highest representable frequency range, down to 0Hz, then back to maximum again. They were generated with Adobe Audition and recorded as 32Bit floating point waves.
Here's a picture of my 'perfect' 22050Hz to 0Hz to 22050Hz .wav file. I.E. the full range of 44.1Khz audio


Now here's what happens when you render it in Ableton at 96K


As you can see, there are many, many artefacts. But wait, there's worse to come.
Here is my 'perfect' 48000Hz to 0Hz to 48000Hz .wav file. Or, the full range of 96Khz audio.


Now here's what happens in Ableton when you render 96K at 44.1K


As you can see, the 96k-44.1k is butchered. Very audible artefacts.. So what can we do to prevent this? Well, you can try only rendering at the rates you record at. Sadly, not everyone has the ability/means to record/process at 96k, and 44.1 is fine for playback but not at all good for processing.
The second option is to resample whatever audio OUTSIDE of Live. The best (free) resampler is Voxengo's R8Brain (http://www.voxengo.com/product/r8brain/). It's an amazing resampler and doesn't cost a penny.