It is a general practice to locate the flow nozzle in the pipe, with 6 - 8 diameters of straight pipe length on each side of the nozzle. 


I have not come across with the impact of inaccuracy due to instrument tap-off within that 6 - 8 diameter pipe length. Maybe I am unaware. 


But we always try to verify the flow nozzle readings with other possible measurements, like, steam glow with BFW flow & calculated B/D water flow etc. 


Since 650DN pipe is a very large one, (26" diameter pipe), my guess is the impacts of Instruments tap-offs should be minimal.  However, there could be several other factors impacting the accuracy of flow nozzles.


Hence, I don't have a clear answer about the % of inaccuracy in a flow nozzle.  May be some Flow Nozzle manufacturers, who deal with 20" to 30" diameter steam piping could give a better answer.