There is this old article by "Saintpid Mastering" (started in March 2016), testing various limiters (both "True Peak" and normal ones that "can" address True Peak through the help of oversampling and long look-ahead times). However, not with a synthetic test, but a mix "boosted" into a limiter and then the 32bit FP (Floating Point) results being analyzed with iZtotope RX.
Here is the chart, last updated about 6-8 months ago:
https://www.saintpid.se/en/isp-true-peak-limiters-test/
Please take note though, that this is an extreme example for "loud material". Also, iZotope RX has different readouts compared to Wavelab 10 (which I use for the statistics sheets). Wavelab is actually a bit more picky.
Another topic (which would break the scope of this thread), is how the limiters behave, and how you should ideally set them up.
For example: Tokyo Dawn Labs Limiter 6 (which is an upgraded version of VladG Limiter No 6). This limiter is built for "keeping things in check", in a quite literal sense. The ISP limiter alone can take care of rogue peaks at low LUFS values, sure... but for really pushed high RMS avg mixes, you need the array. With ToneBoosters Barricade 3 or 4 as another example: due to some internals, setting the ceiling to -1,0 dBTP, and at moderate high LUFS values, can results in the readouts you see on the chart. So a safe solution, for most ISP limiters actually, so to set them to -1,10 dBTP. This way, even if you have overshots, then most analysis tools will still stay in the "green".
Again, big and long topic. In context of the Mix(ing) Challenge however, your should work in such a way that you have a moderate loudness. Considering the -16,0 LUFS ILk (absolute max) value, if you do have rogue peaks reaching over -1,0 dBTP that you need to keep in check with a "safety limiter", either you overdid it with transient designers, or your track is really darn dynamic.