It's a very common misunderstanding. Film speed isn't determined by density because negative density by it self isn't relative to print quality. Film speed is determined by gradient. The fixed density of the ISO standard is a short cut to determining the fractional gradient point. As per Nelson, the fixed density method is only accurate when the contrast falls within the ISO contrast parameters. Otherwise you need to plug ΔD and Δlog-H into the Delta-X equation because with increased and decreased negative development, the fixed density method is less accurate. You're probably conflating it with how an increase in film density usually is accompanied by a higher gradient. This isn't always the case and definitely not to the same degree in different film types. This is all explained in
FWIW, it seems pretty well-established that reducing developing times significantly from standard (whatever that is now, since every manufacturer can decide for itself), for whatever reason (contrast control, personal EI, etc.) will reduce effective film speed. With that in mind, it only makes sense, sensitometrically, that if one develops to a lower contrast index, one's personal E.I. will be lower than ISO speed.