[E within this numerical program] Context: all spectra previously run on this
complex threaded flux only through "bar cocycles" — phases on edges crossing each
removal bar's half-plane, one cocycle per bar. The probe measured where those
cocycles sit inside the true H¹ (b1-formula-exact):
| K | dim H¹ | bars | bar rank | complement |
|---|---|---|---|---|
| 1 | 5 | 3 | 3 | 2 |
| 2 | 81 | 27 | 27 | 54 |
| 3 | 1409 | 219 | 219 | 1190 |
Bar cocycles are exactly flat (max face flux 0.0 at all K) and independent, but span 60% → 33% → 15.5% of H¹. The complement — harmonic 1-cocycles orthogonal to the bar span — is not dead gauge freedom: threading the canonical complement cocycle moves the spectrum smoothly and monotonically (spacing_std of lowest 200 eigenvalues: K=2: zero 2.069 → complement 1.870 → bar-φ 1.150; K=3: 1.716 → 1.611 → 1.465), flat to 1e-10 in phase.
The canonical complement cocycle is 64% (K=2) / 70% (K=3) interior by weight. "Face stratum" was the working name and is a misnomer: the complement is bulk-threading flux freedom that the bar parameterization silently discards.
Connector relevance: this is the construction's native supply of "incomplete
pieces" — more independent flux classes than the obvious menu, growing like
~(20/8)^K relative to the bars. Any mapping of the construction onto particle
content must account for where these classes live (gap-hol3-particle-content-connector).