Closing as this rather is a question about that paper than about ELKI.
To understand the paper, it will likely help if you study the references therein. In particular:
Böhm, C., Krebs, F.: The k-nearest neighbor join: Turbo charging the KDD process. KAIS 6(2004)
Böhm, C., Kriegel, H.P.: A cost model and index architecture for the similarity join. In: Proc.ICDE. (2001)
but also the other references on OPTICS, CPR, and k nearest neighbor joins.
I had read this paper, but it's hard to understand its algorithm logic for me, is there any docs about this algorithm?