We work with both systems and every system on the market calculates yield differently. There are many details: points vs polygons, overlapping of polygons and how to calculate them, how to treat anomalies and 0 values.

What usually works well is to set rules or manually adjust yield datasets with calibration/cleaning (USDA protocol, or AI-based, which calibrates well in case of multiple harvesters).

Just as a reference, Cornell NMSP use-case.

In case you set a bi-directional JD-FV integration, you can also experience acres duplication and data loop issue (we help resolving it).