I start counting from when the water stops visibly / enthusiastically bubbling. And, tbh, that's tended to work. Though having heated equipment is also quite important. Getting everything around the same temperature to start with (or as close as possible) makes things more replicable / consistent. Pouring boiling water into a cold FP might also kinda balance out the effect. Um, I'm aware some people've taken the time to work out how much heat the water loses between the spout and the grinds. But that's getting into 'caring way too much' territory, IMO, and anyhows, they don't seem to dispute the 8ish seconds of the boil thing. IIRC.