The design, realization, and measurement results of a high-accuracy multiyear 3.5 GHz trans-horizon radio propagation measurement system are discussed, with both emphasis on the results and implemented technical measures to enhance the accuracy and overall reliability of the measurements. The propagation measurements have been performed on two different paths of 253 and 234 km length, using two transmitters and one receiver in the period September 2013 till November 2016. One of the paths travels over wetland; the other path can be considered as a land path. On each path, an additional transmitter is placed at 107 km (in the 253 km path) and 84 km (in the 234 km path) from the receiver. With this arrangement, the correlation between two nonaligned paths of comparable length, and two aligned paths of dissimilar length, was studied. The measurements show that for the land path, the predicted ITU-R P.452-16 cumulative distribution function (CDF) typically shows 5 dB higher path loss than the actual measured CDF for the region of interest; anomalous propagation. This means that the measured signal is on average weaker than predicted (a higher path loss). For the wetland path, the actual CDF is very close to the predicted CDF. Also, the measurements reveal that typically 30% of the anomalous propagation occurrences are correlated with other paths.