Looking at the audio trace and I don't think it is Manchester encoded but uses a variable carrier off time with a fixed width on time to denote 0/1 bits. The message start/sync seems to be a longer carrier on/off time.
Can you attach or link to the actual audio file you took? If we can get the on/off timings then knocking up a bit of code to extract the data stream should be easy. Then it's just a case of trying to decode the stream.