The performance of indoor localization systems based on angle of arrival strongly depends on the environment. Consequently, a characterization of localization errors in different environments is necessary to correctly assess the accuracy of these systems. This paper focuses on the localization errors in multiple environments using basic 2.4 GHz linear antenna arrays. First, a theoretical algorithm is elaborated to predict the expected localization error in various environments. Subsequently, this algorithm is tested in practical tests performed in an anechoic room, an empty room and a room with obstacles. In these tests, various techniques for error minimization are evaluated, as well as the Beamscan, ESPRIT and MUSIC angle of arrival algorithms. It is shown that the accuracy in an anechoic room can also be obtained in an empty room for certain configurations. This is not the case for a room with obstacles preventing line-of-sight connections. For this case, a new type of localization system is proposed.