An approach to constructing the static characteristics of a system from experimental data is considered. It is noted that in many cases the problem is solved by applying the "black box" concept, according to which the data of an experiment containing the values of the measured input and output quantities are used. In practice, the input and output variables in the experiment are determined with certain errors. It is shown that in solving the problem within the framework of the conventional approach, the availability of various sources and generating factors of interference models is often ignored, which leads to a significant distortion of error estimates and to formation of an inadequate conversion characteristic. In view of this circumstance, the types and sources of errors appearing in constructing the static characteristic are determined, and the models of noises emerging during measurements under real field conditions and during a calibration experiment are studied, and it is shown that they have fundamentally different effects on the measurement result.