i think i have experienced this. just didn't have an explanation for it. ended up always scaling the rip file by 102.5% (or something) one direction after giving up on the calibration side.
i'll do a few tests to see if a percentage will effect accurately the measurement across the gantry.
still curious, for the record if there is any rastek/efi/daytona/flora horizontal calibration other than bidirectional.
aa