It should be “The frequency specified in calculationPeriodFrequency must divide the regular period precisely”. The regular period of a swap this the time period which is not an initial or final stub period. So the regular period starts on the effectiveDate UNLESS there is an initial stub in which case it starts on the firstRegularPeriodStartDate. And the regular period ends on the terminationDate UNLESS there is a final stub in which case it ends on the lastRegularPeriodEndDate. This is one of the places in FpML where I think the language and structure of confirmations was allowed to impinge upon the data modelling too much. It would be useful to always know the start and end dates of the regular period.
