Figure 3:

Performance of CLAN-based ChiRA compared to naive approach on benchmark data. Being a chimeric read aligner CLAN produced fewer wrong hits and more multi-mapped hits that contain the true alignment. CRL-based ChiRA could pick the correct reference from the multi-mapped hits for any arm length. Note that although there are more multi-hits (green) in naive mode compared to chira mode, the origin of these reads is still uncertain.