28 may jump in earlier on an intrabar cross based on his experience trading the method, but in a backtest you are only cheating yourself if you take the trade early since in realtime we don't know if there even was an intrabar cross and we don't know how the bar will eventually close. You will also never see the intrabar crosses that failed since they will no longer show on the chart. So for the purposes of evaluating a method you can only use a strict interpretation of the rules on the bar close or else you are setting yourself up for disappointment.
Once you are getting the signals correct per the rules, then you can start using discretion in realtime to (hopefully) improve your results.