No. Readings can come from several locations on one shell or repeated measurements of one location. Independent shells, mapped locations and repeat observations answer different questions. Keep that measurement hierarchy visible in the sample records.
For example, ten shells measured at three locations provide thirty location readings from ten sampled shells. Repeating one reading assesses measurement repeatability without adding another shell to the sample.
It depends on the lot, characteristic, decision purpose and agreed sampling plan. Lot acceptance, estimation of lot quality and ongoing process control are different tasks. A thickness study and a machine-fault investigation also need not use the same sampling design.
Define the population, selection method and decision purpose, and preserve the shell/location hierarchy in the results. A sample count chosen for one investigation does not establish a lot-acceptance plan for another.