The initial information on the molecular functions necessary for calculations of spectra is always of discrete nature, being either quantum-chemical calculations or experimental data. Selection of the model functions built on this information and then used to calculate the intensities of the fundamental transitions and low overtones is not restricted by any conditions, yet these same functions can result in errors when used for the high-overtone transitions. In particular, the errors are found in calculations for OH, PN, YO, CaO, PS, NS, SH, PH, and NO. We analyze the sources of the errors and give recommendations of how to avoid them. The molecular functions should be chosen such that the Normal Intensity Distribution Law is fulfilled. In order to increase the accuracy of the calculations, it is also desirable that their analytical properties in the complex plane be as close as possible to those of the real functions.