If only one sampling is measured on the pulsing method, and it is enough to suppress noise, your logic is correct. But actually, averaging (filtering) is needed even on pulsing method.
Averaging (filtering) also has the effect you counted up. And its effectiveness is proportional to the averaging duration (more precisely, square root of duration).
To maintain self-heating as much as DC measurement, pulsing duration should be division of square current.
Thus, as the total, both method give the same performance (S/N) on the same self-heating.

But in the real world, we cannot use entire ON duration information without contamination of OFF duration on pulsing method. Therefore, pulsing method cannot achieve the same S/N as DC method.

Tsuneo