That would be true if you didn't have a fan and a radiator attached to the water. Not all of the heat is going to be absorbed by the water, and the hotter the water gets over time, the less heat is going to be going into the water. If you plotted the temperature of the water over time, it would be an asymptotic curve.To heat 100ml water rise up 20K need 8400J, 140W heat source take 60 secs to provide this power.
The problem if you run system with 100% power in long time is all component temperature will rise, but the test case has a dynamic air flow so air inside keep refreshing any time so result is not different even you run stress test 1h.
Did you actually test it for 1h, or are you just throwing that number out there? Because honestly it'd probably take longer than that.