Filtering ARIMA models based on significant coefficients

Why were the only models with all significant coefficients in this example ARIMA(6,1,3) and ARIMA(5,1,1) if all models between ARIMA(1,1,1) and ARIMA(6,1,6) were tested? Shouldn't all of the significant models found in the previous video "Fitting Higher-Lag ARIMA Models for Prices - Part 1" have been significant again since the same data was used?

Or did this video imply that the ARIMA(6,1,3) and ARIMA(5,1,1) models were the only NEW models with all significant coefficients apart from the ones we already tested in the last step? Thank you!