A laboratory rating is measured between two rooms with no flanking path except the test wall: a heavy, isolated test frame, sealed edges, no services. A site measurement includes everything else the building does — the floor, the ceiling, the adjacent facade, the ducts, and the workmanship.
The three losses that matter most are flanking, penetrations and workmanship. Flanking is transmission around the wall rather than through it: a continuous screed, a continuous ceiling void, a structural column that bridges both rooms. Penetrations are holes that were not sealed, or were sealed with a material that fell out. Workmanship is the board joint that was not sealed, the mineral wool that was compressed into the top 200 mm, the door frame that was packed with mortar.
In a well-detailed building the site result is typically a few dB below the laboratory figure, and that is normal and expected. In a poorly detailed one it can be 10 dB or more below, which is the difference between a quiet room and an unusable one.
The practical conclusion is not to distrust laboratory data but to know what it covers. Specify the wall by its tested assembly, detail the flanking and the penetrations explicitly, and where the acoustic requirement is critical, carry out a site measurement before handover. A wall that is measured is a wall that gets built to the drawing.