At present, the sigma used for evaluating the runs rules violations is always calculated using the input measurement data. It is possible for the user to define the sigma to use for all other calculations (ie. http://documentation.statsoft.com/STATISTICAHelp.aspx?path=QCC/QCCharts/Dialogs/DefiningSetsSpecificationsforChartsfromVariables), but this does not affect the runs rules evaluation (and it should).
I think this is how runs rules use sigma in 13.4