It looks like each cycle takes longer to dissipate the heat enough to ramp back up, though the test cuts off just to soon to tell if the pattern continues. I wonder if it's getting hotter each time, if the case circulation isn't good enough causing the system temp to warm up, making it harder to move the heat away, or if it's something with the throttling algorithm. You said the maximum temp was 93C, which I assume means that's the max throughout the entire test, or was it just the max before it throttled the first time?
Also, while I agree with testing it without a heat spreader since that's how it comes, it should have been tested with one as well to see how it would perform if someone were to add one.
Personally, considering you'll rarely benefit from these crazy speeds, I'd much rather see increases in density/capacity and cheaper prices. I figured they'd be pushing 200+ layers by now, especially since I'm pretty sure I saw something about Samsung hitting the high 200's soon.