There's also the strong possibility that the normal (statistical) process variation within each vendors process is wider than the difference between the two. Thus, even if the median TSMC chip runs hotter than the median Samsung chip, a large fraction of Samsung's chips could run even hotter than either average.
It will be difficult to test, but after a hard reset and then chilling two phones to the exactly same temperature, one could run successive identical benchmarks on both to see which one throttles the CPU first.