Just to let you know I have performed the tests again with the circuit disconnected and achieved considerably better disconnection times.
I was surprised what a difference it makes.
Having checked my own notes I appreciate that it was my error in forgetting to disconnect all loads the first time, which they always should be.
I see GN3 advises it can be dangerous to perform RCD tests with load connected as potentially dangerous voltages could appear on exposed conductive parts as well as extraneous conductive parts.
Perhaps this is why I have heard the argument between people about testing being carried out at the RCD or anywhere in the installation.
If testing an RCCB in a consumer unit for example, this could be done at the board with all outgoing circuit breakers open and thus all load disconnected, which cannot be so easily ensured if testing at a point on a live final circuit.
Thanks for the advice, this was a useful experience for me which I can bear in mind for carrying out testing in similar circumstances at work in the future.