Observing with two separated L-band bands, rtpipe seems to search higher integrations than it should. Candidate found at boundary that looks like it covers data before resampling (dtind=1) when it should stop at limits of resampling (for dtind=2).
Example plot at: http://www.aoc.nrao.edu/~claw/realfast/plots/cands_17A-396_sb33611387_1_1_001.57830.79720269676_sc36-seg0-i836-dm104-dt2.png