You take what you CAN prove and you assign it to the stuff you can't, assuming that constants remain constant and that variables remain um...well constant in their variation.
So, for instance (applying layman examples) I assume that the suction my vacuum creates at one end of the hose is more or less the same as at the other end of the hose, allowing for friction of the surface inside the hose (because I can prove it creates friction in another situation) and the loss due to curvature.
It might not, there might be some random part inside the hose that I can't see or the material might be different somewhere, but I can apply a reasonable assumption that it will until someone can prove otherwise.