It has been proposed to jointly estimate the angles and delays of the paths in a multipath channel. It is interesting to investigate how this improves the precision of angle-of-arrival based localization systems. In this paper, we provide measurement results comparing the performance of ESPRIT-based algorithms for joint estimation to angle- and delay-only estimation in a single and a two-path scenario with different number of antennas and with different maximum length sequences as localization signal. It is shown that in the two-path scenario joint estimation provides improved precision compared to angle-only estimation. This holds true even when reducing the number of antennas used for joint estimation to half the number used for angle-only estimation. Within the results for joint estimation the longer sequences perform best. In the single-path scenario both methods perform similar to each other. There is no improvement in the precision of delay estimation in the two-path scenario.