The measurement provided was simply the "drop rate" per 100 calls. Implicit in this is that it's the drop rate for successfully connected calls, since a call that never connects can't be dropped. I believe that the iPhone 4 is more sensitive and has better reception than the 3GS. There are rooms in my house where I get 0 bars and I can't make a phone call on the 3GS but I can successfully make calls on the iPhone 4.
So let's say I try to make 100 calls, but the reception sucks in some rooms so on the 3GS only 80 calls successfully connect and of those 4 get dropped. That's a drop percentage of 5% or 5 our of 100. But let's say on the iPhone 4 I'm able to successfully connect on all 100 calls, but since the signal is so low in some of the rooms, I drop 10 of the calls. So 10% or 10 out of 100.
So I've now got a drop rate that's double the 3GS, even though I'm able to make 10 more completely successful calls per 100.
Yes, I was shocked. I thought Apple handled this all well, but I was really expecting them to say that the "iPhone 4 still provides better reception than the 3GS". I guess they can't actually make that claim and I still find that a bit funny.