The first thing I notice is the top trace starts and finishes at different levels, the bottom one doesn't.
However, I'm guessing the problem is really that you need to pre-condition the receiver to expect the signal, try sending 0x00 or 0xFF before the real data to allow the receiver to adapt to the incoming signal strength.
These receivers have agc circuits and bit slicers that can easily corrupt the first few bits of a received signal. I agree, send a header of useless bits out first, and then send out the data you really want.
It looks as though the Rx uses a flip flop that changes 0->1 on positive incoming signals and 1 ->0 on negative excursions. Hence RX O/P stuck at a 1 after signal. I would invert the Tx signals modulation.
frank