The electromagnetic field exposure of the population due to wireless communications originates from both down-link and up-link emissions. Although the main contribution comes generally from the latter (e.g., higher by three to five orders of magnitude for the 2G), the former must be considered as well, because they are continual, and as contributions can be competitive for some cases (e.g., in femtocells). Sensor and exposimeter networks (NW) can be deployed by the operators themselves (to enrich feedback information from their own NW) or by independent external stakeholders such as regulatory agencies or local authorities. When sensors are directly worn by a user, body proximity effects – notably the maskingeffect – can introduce significant errors in the ambient field measurement. A methodology of the statistical assessment of this harmful effect is proposed in this article. It is mainly based on electromagnetic simulations (and partly on measurements) of a triaxial sensor – composed of three orthogonal wideband probes devoted to the evaluation of the field components – placed at different positions of a set of whole body phantoms. The main original contribution of the proposed approach is that both the isolated sensor calibration procedure and the assessment of the measurement errors are based on statistical analyses accounting for the propagation environment. The quantitative results are obtained using statistical channel models for polarimetric and non-polarimetric measurements in various propagation scenarios. Some quantitative results examples are presented. Eventually, preliminary corrections schemes are proposed.