The difference between your two scores is 3%. Not sure what the run-to-run variance is, but...
Heat is bad, mkay.
Heat doesn't increase performance.
Repasting likely reduced fan noise (probably didn't thermal throttle at 89C before anyway). But the CPU is still governed by power draw and boost duration.
If you want to get the most out of your CPU, undervolt (which it looks like you did, but have you tested even lower?). This will allow the highest clock speeds for the longest duration.
I'd also recommend maxing out the "Turbo Boost Power Time Window"
Within XTU, you can also adjust the "Turbo Boost Power Max" and "Turbo Boost Short Power Max" settings.
You want to be monitoring CPU frequency throughout the duration of the test. If the score is aggregate (instead of peak), then you're being scored on the full duration of the run. If freuqncy drops off significantly at 28 seconds but the test takes 60 seconds, your score will obviously suffer.
Keep an eye on the throttling indicators at the bottom right corner of the XTU screen also. That will help you to know what to increase to eliminate throttle for the next step.