Completed-Event Transport: Reproducible Record-Access Benchmarks Across Two Quantum Platforms

Testing how retained measurement records determine prediction performance \[\boxed{p(y,r\mid C)=\frac12\langle y|E_r(C)|y\rangle,\qquad \mathcal R_S-\mathcal R_R=\mathbb E[(q_R-q_S)^2]\ge0.}\] How much predictive information survives when earlier detector records are hidden? This two-platform empirical paper connects qualified completed-event transport to explicit joint probabilities, fixed information-deletion rules and measured prediction performance. In the superconducting recurrence archive, retaining four earlier measurements raises fifth-outcome prediction accuracy from 73.2183% to 82.2774%, across all 495 settings and 19.8 million archived trials. The gain over the latest reading alone is 2.4895 percentage points. No outcome probabilities, noise rates, gate angles or contrast normalizations are fitted. All seven record selections, unsupported histories and adverse settings are retained, and confidence calibration is evaluated separately from accuracy. The complete trapped-ion benchmark is preserved: 32,000 tree records from 6,400 hardware shots, five strengths, seven masks and 30 nested comparisons. Its primary calibration p=0.0862 and complementary p=0.0189 are both reported. The two experiments' sampling units and uncertainties are not pooled. The paper is written for scientists new to Quantum Traction Theory. A boxed fourth-face energy-capacity identity explains the proposed source picture; the additional transport, composition and sampling premises needed to construct detector predictions are stated. The archives test the resulting record instruments, not the factor 4 pi or a fundamental length directly. The displayed joint law describes the ion preparation-inference task; a sequential branch law handles recurrence, with the same prediction-loss identity. The matched ordinary quantum instruments have identical probabilities and the same zero outcome-fit burden. These are retrospective, reproducible source-to-record benchmarks, not a claim of experimental separation. A calibration-first follow-up specifies how an independently source-selected apparatus input could add predictive content. The reconstruction ZIP includes editable source, all derived probabilities and scores, checksum-pinned data acquisition, independent score checks and a complete two-platform replay. The PDF is the primary reading copy. Experimental archives: Feng and colleagues and Liu, Tornow, Kessler and Barkai. Source dependencies: Completed-Event Transport, Completed-Event Four-Capacity and Physical Contact Selection. Corpus navigation: DOI Map.