When use frac, the universe of pixels per each class is near to 1 but not exactly 1

[WCMC-vblanque]

All the pixels inside the polygon return a value of 0.99999 but not 1?

We are using a raster with decimal values and a polygon in projection EPSG:4326.

[dbaston]

exactextract is using single-precision floats for the coverage fractions, so we can only expect the sum to be within ~1e-6 of 1