Way way back when I was a power station C&I engineer, I did a course on flow measurement. I've had a look but can't find the notes just now. It was emphasised how much effect a bend or pipe diameter change could have on the accuracy of measurement.
There was a rule of thumb that there was a surprising amount of distance required from a pipe bend/corner/elbow and something like that the flow meter had to be 10x (or was it even 20x?) pipe diameter downstream from any bend BUT ALSO 5x diameter upstream from a bend (ie sensor being BEFORE a bend would also be inaccurate). More than that if you had two bends in quick succession, you need to be even further downstream from the corner (maybe that was where the 20x came in).
Basically for it to be accurate it had to smack in the middle of a long straight, constant diameter pipe. There's about a gazillion types of flow measurement devices all with different susceptibility to this issue, and I don't recall for sure where turbine flowmeters came in but I'm pretty sure they were significantly effected.
So in other words... what IO540 & rans6andrew said
