Because that headline result is not that fewer people died per fire (although it's about the same in both categories—about 1.15-1.2), but that fewer people died period, mostly because there was a lower rate of fires. This means the relevant denominator is "how many people were living in each kind of housing?". To catch up to the same overall death rate, there would need to be 80-90 more fires in newer multifamily homes, which is a few times more than the number there actually were. That seems very, very unlikely to be a chance fluctuation. So if you live in newer multifamily housing, you were considerably less likely to die in a fire than if you live in (newer or older, but especially older) single-family housing.