Hi
Set your sample rate on the scope to something that will get you at least 10 samples per bit at the highest likely clock rate. Look at the data with an IR sensor rather than something that is trying to demodulate it. Grab code samples multiple times. Lay them down on top of each other. You will have errors and this will let you spot them.
After a bit of looking, you *should* be able to work out a likely clock rate. Go out across 10 or more bits once you have a guess to come up with an accurate number.
At that point you can start working out what is preamble and data. Send multiple codes and see what does not change.
Yes, this is generalized / start from scratch approach. It does not use any shortcuts. Because of that, it's got a high chance of success.
Bob