Fluctuations in the Received Signal Strength, caused for example by temporal propagation dynamics or various mobile types, can decrease the positioning accuracy in WLAN-based indoor fingerprinting. In this paper, the effect of an offset between Received Signal Strength values in the training and estimation phases is investigated. Our study is based on a huge measurement campaign that covers in total eight different buildings with several floors each, in two countries. Different offset types and offset values are studied on a large scale, in terms of 3D positioning accuracy and floor detection probability. We will show that biases between −20 dB and +10 dB do not affect the positioning results in a significant way and that such biases could be also tolerated without calibration.