Read Granularity and Positional Integrity in a Deterministic Acquisition Model

Preprint — not peer reviewed. A deterministic file-backed simulation examines how read-error policies affect the content and position of readable sectors. Five 4 MiB references, three fault conditions, three policies, and two technical repetitions produce 90 acquisitions. Each error condition designates 64 of 8192 sectors as unreadable. Sector-sized zero filling preserves all 8128 readable sectors at their original offsets. Zero filling failed 4096-byte requests preserves image length but replaces 448 readable neighbours when faults occupy separate request blocks, yielding 94.488% positional preservation. Omitting failed 512-byte reads retains readable content in sequence while shifting subsequent offsets. Cluster alignment explains the smaller collateral loss under contiguous faults. Set-based predictions agree with every measured preserved-sector count. A fresh execution reproduces all archived CSV rows and logs; this is a technical verification rather than independent laboratory replication. The contribution is a compact, auditable validation example separating positional fidelity, replacement scope, and failed-request coverage. The model does not evaluate physical storage, filesystem recovery, acquisition speed, or the behaviour of named forensic products. The supplementary archive includes the experiment source, original execution records, and a separate verification script and report. The manuscript discloses AI assistance.