We consider an autonomous underwater vehicle (AUV) equipped with a towed hydrophone antenna array which receives signals and echoes from active sonar sources at standoff ranges. Neither the positions of the sources nor the timing of transmissions are known to the AUV. The objective of the analysis onboard the AUV is to estimate these parameters by processing the received acoustic data. The estimate of the parameters is a precondition for the use of the echo contacts generated by the acoustic sources for an automatic target tracking. We call this capability of utilizing stand-off sonar signals for onboard target tracking “non-cooperative multistatic active sonar”. In previous work, we have implemented multi-hypothesis tracking (MHT) filters for automatic parameter estimation, which are evaluating the received direct blasts of the sonar transmissions plus echoes originated by non-moving (fixed) objects at known positions and with point-like target echo structures. It turns out that, as long as the sources keep pinging, the unknown source parameters can be estimated with increasing precision over time, given that the source-objects-AUV geometry is “rich” enough to provide a unique solution of the associated estimation problem. In this paper, we investigate the robustness of this approach regarding ambiguities caused by multiple echo contacts generated by a fixed object with a large spatial extent. In real data gathered from active sonar sea trials, scatterers localized around prominent geographical features (like small islands) are observed. The analysis result for a case study simulation scenario, developed in this paper, indicates that in cases of a sufficient number of point-like objects available, the ambiguities generated by one extended object do not harm the MHT parameter estimation operating with the wrong assumption that this one extended object is a point-like object. If, however, the exploitation of the echoes from the extended object is essential for the MHT parameter estimation to lead to reasonable results at all, then the MHT parameter estimation algorithm needs at least prior information about the extent of this object.