Following on what
telebianchi wrote, if the scale just makes two measurements - conductivity and weight - then it cannot possibly determine accurately five outputs (your total weight, percentage bone, percentage fat, percentage muscle, percentage water). Mathematically, this is what would be called an underdetermined problem. To convert just 2 measurements, it must have some generalized algorithms that determine internal correlations between these supposedly independent variables. Those correlations are presumably taken from models based on laboratory measurements that are good generalizations, but may not apply to you.
HOWEVER, it is possible that the electrical conductivity measurement is not a single scalar, but something more sophisticated, i.e.,something like an oscilloscope, which would give conductivity in a frequency domain. If there were experimental data that showed that different materials (fat, water, etc.) had not only different conductivities, but also different phase response, then what I wrote in the paragraph above would not be accurate. I don't know if the scales are that sophisticated nor do I know that the materials have conductivities with distinct frequency dependence. Maybe one of you know.