As long as blood pressure variance is randomized then you’re getting a signal, and that happens when everyone is measured in a similar way (or randomized to different ways). You don’t need perfect precision.
What you don’t want is to make everyone who, say, smokes wear the cuff while standing up while everyone else gets cuffed lying down.
> and that happens when everyone is measured in a similar way (or randomized to different ways)
But that's exactly the issue -- that the similarity or randomness is one way for one study, and another way for another study, because of culturally different sets of nurses and/or patients.
In other words, you're right it's not affecting results within a study, but it makes comparison between studies questionable.
And it makes it equally questionable whether a study's results apply to you, if your signal is 10 units off of a study's signal, and occurs across a cutoff that determines whether you should take a medication or not.
What you don’t want is to make everyone who, say, smokes wear the cuff while standing up while everyone else gets cuffed lying down.