Totalizer In Micro850 PLC

What lengths of straight pipe are upstream and downstream from the flow meter to the next bend or valve?

What causes the pipe to not be full? Would a check valve help? Anyway, If the pipe is empty then that is the probably the best candidate for the accuracy problem.

Also, if the pump is off and on, e.g. if it is filling a tank using bang-bang control, then that means the 20kgal per day is accumulated at a much higher rates, and the low flows I mentioned earlier may not be an issue.

I would say it's around 10-15' of vertical pipe before a bend. here's a picture of how everything is installed. I would think any excess water in the pipe goes back into the well after it stops running.

If those flowmeters' probes, with the paddles on the end, extend significantly into the pipe, then that increases the velocity of the water by (i) decreasing the cross-sectional area of the pipe at the meter and (ii) by moving the paddles away from the wall of the pipe into faster-moving water, both of which will artificially increase the reading, although that second effect will be small if the Reynolds number is high enough.

Are those white goosenecks on top vacuum breakers, with check valves, perhaps?

I suggest doing a bucket test to check the calibration of the meters.

According to their website they do have a pulse output as well as a flow sensing output, so I suggest you look into this option.

https://www.omega.co.uk/pptst/FPB100.html#order

There is no doubt that trying to use a linear signal i.e. 4-20ma is not going to give the same accuracy as a pulse, given a one second snapshot the flow could vary a little, especially if the medium contains solids or varies in density, over time even an hour this will give quite a significant difference.
There is probably no need to have very fast pulses perhaps set the pulses per gallon to 1 i.e. only measure whole gallons, the nature of these units means that the unit itself will keep the pulse output in line with the actual flow at any given moment.
If as the OP says, 20000-30000 gallons a day then that would only be 20 pulses a minute, so even at 0.1 gallon setting that would only be 200 pulses a minute well within the scan time of most PLC's without using high speed inputs or interrupts.

How many times per day does each pump stop and then start again?

Speculation follows:

If the flow stops and starts, and when it does, the water in those risers empties, then some volume of water measured by, and that has been pumped past, these FPB151s will never reach the downstream process (tank?), and instead drains back into the well to be measured again. Not only that, but as it drains it could drive the paddles backwards which might generate pulses and be interpreted as flow.

Assuming 3.8" ID, one gallon is about 20.4" of pipe. Assume 2gal above each flowmeter (including the horizontal run) on average, with those 2gal measured twice i.e. down and up, every time a pump stops and starts, without leaving the well.

If each of the six pumps stops and starts every 3.5minutes on average, that would "measure" an extra 9874gal/d that would not be measured by the flowmeter on the upstream side of the well.

Again, this is pure speculation or maybe fantasy; I have no idea of the actual characteristics of this process e.g. the pipe ID, or the volume of water that is above each flowmeter that returns to the well, or if the paddles would generate a flowrate when running backwards, or how often the pumps stop and start, or the scheme of how the pumps start and stop. So this could be a completely implausible and ridiculous scenario: 2500 pump stops and starts per day sounds like a lot to me.

I agree with DR, it will depend on a lot of factors, for a flowmeter to be accurate the pipe needs to be full all of the time & this is usually achieved with a shutoff valve before the flowmeter or at the very least some way of keeping the flowmeter full & no movement.
Should the pipe empty or the fluid causing a sucking or pushing motion in the pipe will also cause fluctuations in the flow total, again, if the pump is starting & stopping quite frequently then this will cause the same problem as above should the pipe have at the very least air bubbles in it.