Terry is probably right... for differences that small, you might need spatial analysis tools.   However, it looks like the tools you are using might have a limited number of ranges.   You have some pretty significant color changes between adjacent sites, and that makes it hard to pick out trends.    We use a visual gradiant theme that has 100 breaks, and tends to provide less color variance between adjacent sites.  Of course, if your yield data is noisy, that won't make much difference.

The FOViewer doesn't let you edit themes, but it uses one of those 100-break themes that ranges from +/- 2 standard deviations from the mean.  Have you tried it?   Does it make the change more visible?    If you want, you can send us the raw data file of this field, and I can create a gradiant theme that covers a much smaller range, and see if that helps the visual analysis.