MAX17048 rough calibration for Panasonic NCR18500A (similar to NCR18650A)

I'm using the MAX17048 in an application with the Panasonic NCR18500A lithium-ion cell (chemically very similar to their classic NCR18650A, I believe) and getting very poor SOC estimation performance.

I'm not equipped to do a full battery characterization as suggested in the data sheet / app note (I don't have a thermal chamber, for starters) but I have some basic data. See the following rundown logs (apologies for the gap in data-- I accidentally put my data logging laptop to sleep-- I'm doing another rundown now):

As you can see, this application has a fairly low rate of discharge relative to capacity-- around 0.05C (20 hours to full discharge), so the measured voltage probably isn't far off from the open circuit voltage. You can also see that the SOC estimation (blue line) is a very poor underestimate of actual SOC, it reaches near-zero less than halfway through the actual discharge cycle.

In a charge cycle, it does slightly better but is still not great:

This is charging at a continuous rate (~0.8A until 4.2V is reached)-- initial SOC estimates are again a great underestimate; after 3h it only shows ~10% SOC, even though the battery is probably closer to 50% charged. I'm thinking this is because the default parameters of the ModelGauge are calibrated for a different voltage profile. At 3.6V, this battery still has half its charge left, but the MAX17048's ModelGauge estimates SOC at around 10%.

Again, I don't require high precision modeling (I'm OK with 10-20% error, even), but it needs to be better than this, I could probably do a better SOC estimation by licking the electrode! (Don't worry I'm not licking the electrode.) So, some kind of rough suggestion about what to set the TABLE registers to would be helpful. For reference, here are related posts from people with this specific issue (poor out-of-the-box performance on 1S lithium-ion non-polymer cells):

• "The slope of the SOC is large" (note, I think adjusting RCOMP alone won't help as there's a basic voltage/SOC mismatch)
• [StackExchange] "MAX17048 wrong SOC readout"
• [Arduino forums] "MAX17043 giving weird readings" (note, MAX17043 not MAX17048)
• [Particle forums] "Particle Boron MAX17048 | Battery Capacity Calibration?" (note again, MAX17043)

Post about SOC estimation being wrong under other circumstances:

• "MAX17048: State of Charge is not correct when battery is fully depleted and board is powered on by the charger"

General questions about how to adjust the MAX17048's calibration

• "TABLE Registers (0x40 to 0x7F) MAX17048"
• "Max17048g fuel gauge discharge rate meaning"
• [StackExchange] "MAX17048 - battery characterization data feeding"
• [Adafruit forums] "Voltage of supported lithium polymer or lithium" (note bad advice)