Ok I made some progress in narrowing down the problem.
First, there was some loose connections with my apparently not so l337 soldering skills . . . after figuring that out, I then did a loop back connecting the TX and RX pins on the SF TTL to USB converter CP2103 (gamefreak, this is what Im using to convert the signal).
No matter what I transmit, nothing gets received. So I scoped up the TX line (see first attachment). Yesterday it looked worse, but today you can still see these rounded edges when it shouldnt . . . You can also see that its transmitting at a baud of 20000. The data sent is the letter 'H'
Since Im bypassing the easy radio that means that it isnt part of the problem (well, it could be, but it wouldnt be the only one). I suspect my SF usb thing is broken and I should just buy a new one . . . thoughts?
Now scoping the UART of my ATmega644, I get the second attached image. The square waves are nice clean and sharp. You can see that the baud is also 20000. I think I was sending 'r' but I forgot . . . Unfortunately after sending this clean signal directly to the RX of the USB converter it still doesnt receive data, making me suspect the RX is just fried entirely.
Lastely, I found that if I removed my divide by 8 (and a few other minor code changes) that I can get a baud that is exactly 19200 coming out of my UART. The datasheet says I will get a 7% error if I leave my clock divided by 8 - think this will cause a problem? I'd rather leave it divided . . .