Yes, but far fewer ACKs come from the sender (there's no additional ACK for a received ACK), making it far less of a problem. You can test it yourself, you know, as I have done, myself; simply run many connections simultaneously. I used to have to do so when downloading stuff via HTTP from the BRTURBO server farm, for example, as many as 150 separate downloads using 4 connections each simultaneously; it's not just BitTorrent that's affected. This will, of course, generate the largest amount possible of TCP/IP overhead, including ACK packets.
When you do this, you'll find that flooding your up pipe causes a VERY significant falloff in download speeds, but throttling the down pipe causes no corresponding change in the up pipe. Ergo - as makes sense - throttling your downloads does little, beyond preserving bandwidth for other network use. Your downloads are self-limiting; if you "go too fast", they simply...won't go that fast. Think of it this way: You're essentially saying "My downloads are going too fast; they're going to slow down my downloads!".