The 48 bit and 7 bit bxing numbers for R7134010 do not agree:

Here’s a good run from F7853 — x-axis is bxing number, y-axis is 8-bit spin assignment, z-axis is # of events:

Here’s the spin pattern using the 7bit bxing for the bad run — it’s ~identical to the one above:

and here’s the pattern using the corrected 48bit bxing — notice the large number of counts in the abort gap #31:

Conclusion

The offset in the spin DB used to correct the 48 bit bxing number for R7134010 needs to be shifted by 1.