Thanks for sharing all relevant input and output up front!
Maybe this is just a numerical issue. In principle we are of course just counting the number of states at the Fermi level in both cases, so they should agree. (We can see this very clearly from the k-point resolved graph, where the band edges give rise to sharp jumps.)
But the algorithms used in the two cases are quite different, so sampling a single point that might be a sharp peak could mean the results differ numerically but on the whole, if you computed the transmission for several energy points in an internal around the Fermi level, they might look similar qualitatively (and integrate to the same current).
The old algorithm for bulk transmission required the system to be repeated in C until it was as long as an electrode should be. The new algorithm I think doesn't need this, and similarly we now can use minimal electrodes and instead set the electrode copy to be sufficiently "deep" to capture all interactions. That is why the dashed line contains two repetitions. Now, that mean 8.8 Angstrom which still is a bit on the short side... You could retry the device config with the electrode copy length set to 3x the electrode.
Finally, k-points work a bit differently in bulk and device. You took care to use the same density etc, and it's a high number, so I guess it will not make such a drastic difference.
Finally, you might want to double-check that no disturbance of the coordinates was accidentally made to the device after it was built. Sometimes when you click to select some atoms, they move a tiny bit not seen by the eye, but if the Undo button lights up, so know some change was made unintentionally.
