Back at looking at this again after the weekend, and I thought I would knuckle down and get my brain in gear to try and understand how the resistance values mikerj arrived at for the differential op-amp circuit were calculated.
Once it twigged about the relationship between slope of the line and gain (i.e. they're the same thing - and mikerj basically said so already!

), and that I could calculate the slope trivially using a spreadsheet function, I thought I would work out exactly what the gain figure should be for the differential amplification. It worked out to be a slightly different value: 1.258, versus the 1.277 mikerj estimated. Using this new figure in a diff. op-amp circuit calculator, I came out with resistor values of 120k & 150k.
However, simulating the circuit with these new resistance values actually gave a slightly greater error than the original. However, the, err... parallelism? of the line on the graph matched that of the original sensor's much better; the old values actually had worsening error towards the bottom of the pressure range (as much as 5% at 0.25 Bar). All it appeared I needed to do to get a better match was to bump the output up a tad, and so I guess this is where the offset voltage divider comes in. After playing around a bit, I settled on changing the 6.8k to 7.5k, which netted me <1% error across the board.

Here's a graph of the results:

I'm still not clear about how the offset voltage figure was arrived at, though. Is it simply the distance on the graph between where the original sensor intersects zero on the Y-axis and the Y-value of the same horizontal point for the GM sensor? That appears to be around 0.15V, which matches what a 6.8k/220R divider gives when supplied with 5V. But the 7.5k/220R I experimentally ended up with is kind of away from that, at 0.142V...
