Thanks Sally and Charles
I ran a series of delete queries on the database that removed the survey and all its records.
In re-importing the records I also got to the bottom of what was causing the invalid location records, as I reported on this forum recently. In the dragonfly Recorder dataset I treat all vice-counties as sites in the location hierarchy, allowing me to easy export records to vc recorders when needed (through the location hierarchy or rucksacks). All sites and locations are currently treated as locations stored in the location name field of the sample and survey event tables, linked to their specific vice-county stored in the location hierarchy. When importing the records, using the import wizard, problems arose in multiple locations within sites where the locations were in differing vice-counties (ie the site crossed the vc boundary).
This meant that within a survey event there were multiple samples linked to differing vice counties in the location hierarchy (effectively multiple sites within a single survey event). Consequently on export Recorder made these samples invalid. All that was needed to stop this happening was to add a vc number to the offending location names prior to import, making them different and so preventing them from being lumped together into one survey event.
Fortunately this problem only arises infrequently, but is one I will need to look out for in the future.
Best wishes
Graham