If you can't do it digitally, you can't guarantee your totalizer will exactly match the display on the meter.
Ours is now within 1% over a 24 hour period with flow rates from zero to 500gpm with the addition of interrupt routines to take fine resolution snapshots of perfectly scaled and calculated ADC data, and we even threw in the measured variation of the execution of the routine to get there, but still, 475,673.1 <> 473,867.6 to the Engineering Manager.
Go digital, or write the code design spec. including the actual variation. People think that perfectly calculated and calibrated high resolution is enough to totalize GPM into G, but they don't seem to realize that the M is critical to accuracy, so how do you communicate the Time in the original GPM signal so that it can be correctly divided on the other end?
You can't.
Go Hart.
Go serial.
Move the wires to DC I/O and reprogram to count and totalize pulses. (that's a right now kinda "make it exact" answer usually at only the cost of your time).