The time scale is a little different. The captured signal is a little faster than the documented 500uS. I doubt if it matters, and if it does matter the difference might be a timing inaccuracy in your capture of the original signal.

The bigger difference I see is the checksum of the repeated part. The documentation very clearly states that R "must be included in the checksum calculation". But in the captured signal you just quoted, R is not included in the checksum calculation.

With that change, the .irp would be:

Device=79
Function=0..7
Frequency=38100
Zero=500,-1000
One=500,-500
define X=(D+F)^1
First Bit=MSB
Form=7000,-2800,0:1,D:8,F:8,X:8,500,-60m;7000,-2800,1:1,D:8,F:8,X:8,500,-60m