The bias errors of transmission tube measurements are evaluated using an empty test tube condition, implying full sound transmission, normal-incidence sound transmission loss (nSTL) = 0, and two narrow tube elements presenting a theoretically known sound transmission loss (which includes modeling of thermo-viscous losses), varying in frequency around moderate non-zero nSTL levels, and approaching insulation levels more typically found in applications. Results show that the different reference conditions are virtually equivalent in presenting negligible amounts of bias error within the corresponding measurement uncertainties, typically within ±1 dB, in the full 6000 Hz measurement frequency range in this particular case. Slight differences are evident only in the upper third of this range, from 4000 to 6000 Hz, in which bias errors for the narrow tube elements are found very slightly beyond the measurement uncertainties, suggesting the existence of additional sound loss mechanisms. Measurements of a typical insulating test sample are also presented for illustration of the possible non-significance of bias errors in practical cases.