Why a pipette always leaves some liquid behind
Retention, and why the delivered volume is never the drawn volume.
A pipette always leaves some liquid behind because the inside surface of its tip attracts the liquid it just carried more strongly than the liquid wants to let go of itself, so a thin film stays wetted to the plastic even after the plunger has pushed everything it mechanically could out through the opening.
A film held by the same pull that drew the liquid in
Whether a liquid clings to a surface or beads away from it depends on a contest between two pulls, how strongly the liquid's own molecules attract each other and how strongly the surface attracts the liquid, a relationship known as wetting. A pipette tip is made from a plastic that wets fairly readily with water-based liquids, since a tip that beaded liquid away entirely would be useless for drawing anything up. That same wetting refuses to let all of the liquid back out again. The film coating the inner wall is held by the attraction that pulled it in, and the plunger's push has no direct way of overcoming it for a layer only a few molecules to a few micrometres thick.
Dipping a spoon into a jar of honey and lifting it out never returns the spoon bone dry. Honey coats the submerged surface, keeps dripping for a while, and even once the dripping stops a visible film remains that no amount of shaking will remove. Nobody expects to recover the exact volume of honey the spoon displaced. A pipette tip does the same thing on a scale small enough that the film is invisible, which is why the loss reads as a mysterious instrument error rather than familiar honey-spoon behaviour. The spoon also shows why the plunger cannot help: pressing harder on the honey still in the jar does nothing to the film already stuck to the metal, and pushing harder on the air behind a pipette's liquid column does nothing to the film already stuck to the tip.
Tip shape, liquid and finish
Retention scales with how much wetted surface the liquid was exposed to relative to the volume being moved. A narrow, long tip presents more wetted wall per microlitre than a short, wide one, which is why a tip designed for a smaller working volume is shaped differently rather than being a scaled-down copy of a larger one. The arithmetic is simple enough to see directly: halve a tube's diameter and the wall area per unit length halves, but the volume it holds per unit length falls to a quarter, so each microlitre in the narrower tube touches twice as much wall.
The liquid matters just as much as the tip. A more viscous liquid clings in a thicker film, and a liquid with markedly different surface energy than water (an alcohol, a detergent solution, an oil) can retain a very different amount through an identical tip, so a pipette calibrated against water can be meaningfully wrong the moment it moves something else. Surface finish plays a part too: a slightly rougher moulded bore offers more real contact area for a film to cling to than a polished one of the same nominal diameter, even though the two would look identical on a drawing that only specified diameter.
A loss that repeats instead of averaging out
Because the same tip and liquid combination loses a broadly similar amount on every dispense, retention is systematic. A random error shrinks when many measurements are averaged; this one simply repeats, so ten dispenses each short by the same amount add up to ten times the shortfall. A loss that would be dismissed as noise at litre scale becomes a real, repeatable error at microlitre scale.
That same regularity makes it correctable, and most precision liquid-handling work leans on this fact rather than trying to eliminate the film. Dispensing a small excess, calibrating a specific tip-and-liquid pairing against a known reference, or reverse-pipetting (drawing up more than is needed and dispensing only the required portion) all work around a loss that can only be accounted for. Treating the drawn volume and the delivered volume as the same number is the most common beginner's mistake in this kind of work.
Air cushions and pistons
Retention is worse in an air-displacement design, where the plunger pushes a cushion of air that in turn pushes the liquid, because the liquid's leading and trailing surfaces are both free to leave films behind with nothing physically forcing them out. A positive-displacement design, where a piston sits inside a disposable capillary and pushes directly against the liquid, cuts that loss substantially, since far less wetted surface lies between the piston and the exit. The capillary walls still wet, so some retention remains. The designer has shrunk the surface area available to the film, and the physics carries on as before on that smaller stage.