I did these manually, and agree that some learned codes are the same. However, looking at the 2 digit pairs, it seems that the pattern is to send the digit or digits, and follow that with Function 31. That's the pattern Jason shows, but generalized to show how multiple digits are handled. So probably the overall protocol is not exactly a 2 part protocol, but rather as many Pioneer signals as digits plus a terminating signal with function 31.

From looking at all the data, I guess that the function numbers are:
0----0
1----7
2----10
3----88
4----3
5----6
6----80
7----1
8----14
9----93

I don't know why some single digit learns were the same, but I wonder if in some cases the learns were mis-labled.

Can you learn some longer sequences of digits? That will allow us to see if the above hypothesis is correct, and to have a little more data to sort out the function codes.