Absent is an attractive word to write into a specification and a bad one to rely on. No sorting, washing or detection system removes one hundred per cent of anything, so a clause requiring absence is breached by a single fragment in a twenty-tonne consignment.
What happens next is predictable. The buyer finds one stem, does not want to reject a whole load over it, and accepts the lot. The clause has now been waived in practice, and the next time something more serious appears the supplier will point at the precedent.
The workable structure separates two categories of contaminant.
Soft extraneous vegetable matter – leaf, stem, cap, strig – gets a percentage tolerance by weight, or better, a count per unit weight where the item is light but highly visible. A strig weighs almost nothing and passes a weight-based tolerance easily while being immediately obvious in a clear retail jar.
Hard foreign matter – stone, glass, metal, wood, hard plastic – gets a zero commitment supported by a described process rather than by a number: the detection technologies in sequence, the critical control point they sit at, and the retained rejection records that prove the system runs. That is a process guarantee, and it is the only honest way to write it.
The same logic applies to microbiological specification. Regulation (EC) No 2073/2005 expresses criteria as sampling plans with defined sample numbers and limits precisely because absence is a statistical statement about a sampling plan rather than a property of a lot.