Appendix: source scope and release-mechanism fit

For each narrow primary statement, retain publisher, title, edition or access basis, locator, jurisdiction, limits and retrieval receipt. The receipt identifies retained bytes or browser evidence; it does not validate every interpretation. The claim ledger must connect the sourced note to its card and locator so an independent reviewer can examine the actual support rather than inherit authority from retrieval alone.

Institutional sources retain their original purposes. Government project guidance concerns government roles and controls. DOE material concerns its project-management context. GAO guides supply assessment methods within their stated scope. World Bank and EPEC procurement or PPP guidance have specific institutional and public-sector boundaries. Service-development guidance concerns technology and public services. Their inclusion does not make them universal law or a contractual requirement for an unnamed private project. The paper uses them as scoped methods rather than as authority to assign real obligations.

Fictional duration ranges are authored mechanism inputs, not source-derived rates, project populations or probability calibrations. Their computations come only from the canonical input and compute contract. A publisher's methodological support does not validate those parameters. Keep each numerical use labelled illustrative and bound to its frozen study so source authority cannot be mistaken for task-estimate evidence.

The practitioner records, objections and qualitative cases are proposed original reasoning, not reported engagements or observations. Test transfer through actual scope, consequence, authority and evidence. Applicable instruments and competent specialists determine material technical or legal requirements. Publication supplies no mandate, professional certification, recipient acceptance or legal opinion; it supplies a structure for identifying the real undertaking's unresolved questions.

The bibliography supports verification without repeating source summaries at length. Narrow sourced notes stay adjacent to their analytical use. The retrieval archive remains part of the production evidence, while public links allow readers to inspect the publisher’s material. A publisher page can change after the evidence cutoff; the retained receipt records the basis used in this draft. Later editions should review both current source status and the manuscript claims affected by any change.

An approval event can be represented as a task only when its meaning and timing basis fit the network. A prepared submission, a reviewed submission and an authorised release are different outcomes. The predecessor should release what the receiving activity actually needs, while the duration should represent the remaining process under its stated conditions. Treating a document as approval because the file is complete confuses production with authority. Treating every approval as a generic delay can conceal the review window or conditional resubmission mechanism that governs the next usable release.

A bounded review activity with supported continuous elapsed duration can belong in the supplied finish-to-start network. The authorisation event remains the task outcome, and the estimate basis should identify required inputs, scope and responsible review authority. This representation is useful when the review starts after its predecessors and its duration can be meaningfully declared under the accepted conditions. It does not prove that the authority will approve the proposal. A network date and an admissible decision are different properties, so the commitment record should state the condition that governs use of the forecast.

A review available only in scheduled windows has another timing mechanism. Completing a submission just after a window can create a much longer wait than completing it just before. A continuous-duration estimate does not reproduce that discontinuity simply because its average seems plausible. The team should retain the submission cutoff, actual review calendar, entry criteria and treatment of a missed window. If the window materially governs the commitment, use a calendar-aware representation or an explicit supported scenario outside the current demonstrator. Do not relabel the app's elapsed finish as a booked authority date.

Conditional rejection or resubmission creates a further boundary. The current app requires an acyclic network of supplied activities. It cannot loop through an unknown sequence of review, redesign and resubmission until an authority accepts an outcome. A team can define a bounded scenario containing a specified additional review sequence, but its inputs and interpretation must remain explicit. It should not disguise an adaptive process as a closed deterministic baseline. The decision may require a staged commitment that separates initial submission from later authority-dependent work, preserving what can actually be authorised now.

Scarce review or inspection resources can govern concurrency even when precedence is clear. Several tasks may be ready together, while the same person, facility or access arrangement can serve only a limited set. The app's earliest dates assume that supplied precedence is sufficient; it does not allocate those resources. Additional precedence can sometimes represent an already agreed sequence, but choosing that sequence is an operating decision with evidence and authority. A planner should not add arbitrary links merely to make the dates look feasible. Where resource competition matters, use the relevant allocation method and preserve the basis of the chosen order.

These timing distinctions matter to downstream value. A supplier mobilisation, financing condition or benefit-start assumption may rely on an approved operating event. If its input instead uses a document-ready or technical-test date, the bridge can be wrong even when the DAG calculation is exact. The receiving model should name the event it consumes and identify the evidence that establishes it. A timing model should not turn a technical forecast into a legal permission, and a financial model should not turn a forecasted permission into an executed entitlement.

The review decision has several defensible routes. Correct the activity definition when the wrong event was supplied. Correct the network perimeter when a required obligation is genuinely omitted. Use a calendar or resource method when the mechanism lies outside the continuous-time DAG. Define an explicit bounded scenario for a known additional sequence. Hold a downstream commitment when the authority or operating evidence is unresolved. The useful route is the one that changes the decision object accurately, not the one that produces the earliest completion number.

The selected-path normal reading does not repair any of these missing mechanisms. It uses the supplied ranges on the selected critical path and excludes competing-path uncertainty and dependence. A percentile-like percentage cannot compensate for a misdefined approval event, a missing calendar or a shared resource. Before discussing confidence, establish whether the model represents the process whose finish matters. The optimistic and pessimistic cases remain declared input contrasts, while the relevant calendar, authority and operating records govern the interpretation in an actual undertaking.