Subsets of four boutons from one mouse neuron allowed 0–12% delayed release within 1% efficacy and 0.01 pulse index limits under linear sums with equal quantum weights

We tested what a structural input count can specify about release timing when functional edge summaries are also given. Source tables provide 137 boutons from one cultured mouse cortical neuron. A linear model assigned distinct boutons to a target with equal positive quantum weights. Total efficacy stayed within 1% of a reference. The efficacy-weighted pulse index stayed within 0.01 of a second reference. Among all subsets of four boutons, 20,092 met both limits. Their delayed release fractions ranged from 0 to 11.9994%. Two disjoint subsets attained the bounds. Complete enumeration checked all 14,043,870 four-bouton subsets. Tighter limits of 0.1% and 0.001 retained a 0 to 9.2297% range. Moving every recorded fraction halfway toward the pooled source fraction changed the main range to 1.6070 to 7.6067%. This is a stated model change, not a noise correction. The source data do not identify target assignments. Recorded estimates are fixed model inputs; 71 recorded zeros do not prove zero underlying release probability. A separate nested test on 60 group means from 20 neurons found a 10.42% error decrease after adding the pulse index to efficacy features. Whole-label refits gave a wide resampling range that included worse prediction. The released preparation labels also differ from the article count. The contribution is a conditional allocation bound, an audited dataset, a reusable solver and a prediction benchmark. It does not measure a connectome or infer actual target responses. The package preserves source data, provenance, code, all results, figures and offline repeat instructions.

synaptic release · bouton allocation · connectome models · asynchronous release · short-term plasticity · fractional optimization · mouse cortex