In response to inquiries from readers of the paper, we have reexamined the time series recorded at DART 32067 and confirmed the existence of a data gap in the recording between 19 and 34 min after the earthquake initiation time. The recorded time series also includes short-period ground oscillations and long-period tidal fluctuations. We originally had run the records through a script that re-sampled the data to uniform intervals and filtered out the tidal signals. This inadvertently interpolated across the data gap and distorted the signal. We misinterpreted the interpolated feature as the second arrival of the recorded tsunami in the top panel of Figs. 3e, S3e, and S4e. The recorded waveform in the corrected version of Fig. 3e appended herein has instead been de-tided using the OTIS Regional Tidal Solutions (http://volkov.oce.orst.edu/tides/PO.html). The initial arrival, which is visible in the original and corrected time series, correctly guided our iterative procedure to match the arrival time but the amplitude and precise shape of the main signal cannot be confirmed. Our preferred fault model provides good agreement with the recorded tsunami waveforms at La Libertad, DART 32411, and DART 32413 as well as with the seismic records and geodetic measures, so exclusion of the interpolated signal at DART 32067 would not significantly change the fault model. This is indicated by the sensitivity of the other computed tsunami waveforms in Figs. S3 and S4. As such, this processing error does not significantly affect the results and conclusions of the paper other than mildly increasing uncertainty in the absolute placement of the slip along dip.