I think you mean "discrepancies between body weight predicted by measured body composition and actual body weight".
Because body composition is generally measured as fat, and “fat free mass”, which is a theoretical tissue composed of 72% water, 21% protein and 7% mineral by weight. But not all FFM is the same. Bones are obviously more mineral, and mineral density is 3 kg/l, while protein is about 1.3 kg/l and water is obviously 1 kg/l. Fat is pretty much fat... 0.9 kg/l. So if someone has big bones (literally), they'll be denser and heavier.
Even if you knew a body's fat percentage exactly, weight calculations of FFM would still vary by about +/- 4% because of the aforementioned differences between individuals and the exact composition of their FFM. However, the only way to truly know someone's exact fat mass is to kill them and use a solvent to extract all their fat. Intramuscular triglycerides won't show up on a skin caliper test, might be masked in a DXA scan, but will definitely be accounted for if using an air or water displacement type of density measurement, but the density measurement itself is based on the theoretical FFM composition. So inaccuracies in the various measurement techniques will also contribute to variability in any bodyweight calculations based on those measurements.
