The problem isn't recovering the recorders when they're underwater, it's finding them (there's always some sort of ROV or submersible that can get to the wreckage). So all the battery power available should be used for the pinger (beacon), not for data transmission.
Also, the data rates for sonar data transmission are really, really low (and we're talking sonar, not ultrasound, which is much higher frequency, but won't travel more than a few feet). Like hundreds of baud, if you want to be confident that the message has been received at the surface. It would take tens of hours to download the CVR/FDR contents using acoustic means (think: huge amounts of battery power), and there's a good chance you'd lose some data; and some significant percentage of the time, the complex data transmission system wouldn't work at all (vide the difficulty in just finding the recorders from AF447). It's just not a cost-effective way of spending money on safety.
Same thing applies to RF transmission - you're making a whole bunch of assumptions about where the recorders are & that the data signals can make it to the overflying aircraft. There's no location that I can think of where it wouldn't be less problematic to just pick up the boxes. You're always going to put some sort of crash investigation team on the ground, anyway. Use battery power for RF locator beacons, not data transmission.