The common kernel is a twist of an old restriction kernel
Fix a finite-generator Coxeter system with symmetric coefficient algebra \(A\), a simple reflection \(s\), and a tail word \(\mathbf t\) with compatible-tuple lifting on upper sets. Let \(U\) be upper, put \(J=U\cup sU\), \(K=U\cap sU\), and use the two-stage kernel data \[ D=\ker(L_{s\mathbf t}(U)\to L_{s\mathbf t}(K))\cap\ker F_U, \qquad T=\ker(L_{\mathbf t}(J)\to L_{\mathbf t}(U)). \] Then the injective second-coordinate maps \[ \gamma:D\to{}^\tau L_{\mathbf t}(sU),\qquad {}^\tau j:{}^\tau T\to{}^\tau L_{\mathbf t}(sU) \] have equal images. They therefore induce an \(A\)-linear isomorphism \(D\simeq{}^\tau T\).
Proof
Step 1. Lift an element of the common kernel
Take \(z\in D\). Choose \(m\in M_{s\mathbf t}\) with \(p_U'(m)=z\). The two defining conditions of \(D\), together with the restriction and first-map formulas, give \[ p_K'(m)=0,\qquad p_U(\mu(m))=0. \]
Step 2. Patch its second component to a tail vector
The zero-overlap version of patched-tail lifting applies to \(m\). It supplies \(n\in M_{\mathbf t}\) matching \(\mu(m)\) on \(U\) and matching the underlying \(\nu(m)\) on \(sU\).
Since \(p_U(\mu(m))=0\), the matching values on \(U\) give \(p_U(n)=0\). The old vector \(d=p_J(n)\) lies in \(T\), and its restriction to \(sU\) is \(p_{sU}(\nu(m))\). Thus \[ \gamma(z)={}^\tau j(d), \] where \(d\) is now regarded as an element of \({}^\tau T\). This proves one image inclusion.
Step 3. Insert a lift from the old kernel
Conversely, take \(d\in{}^\tau T\). Choose \(n\in M_{\mathbf t}\) whose projection on \(J\) is the underlying image vector \(d\). Since \(d\) restricts to zero on \(U\), we have \(p_U(n)=0\).
Set \[ z=p_U'(1\otimes n). \] Its first image is \(p_U(n)=0\). The seed-vanishing theorem gives \(p_K'(1\otimes n)=0\), so \(z\in D\). Its second image is the twisted \(p_{sU}(n)\), which is exactly \({}^\tau j(d)\).
This proves the reverse image inclusion.
Step 4. Use injectivity to obtain the linear isomorphism
For each \(z\in D\), Steps 1–2 give an element \(d\) with \(\gamma(z)={}^\tau j(d)\). Injectivity of \({}^\tau j\) makes that \(d\) unique. Define \(\Phi(z)=d\).
Linearity follows by applying \({}^\tau j\): the image of \(\Phi(z+z')\) equals \(\gamma(z+z')=\gamma(z)+\gamma(z')\), so injectivity gives \(\Phi(z+z')=\Phi(z)+\Phi(z')\). The same argument proves scalar linearity.
Step 3 and injectivity of \(\gamma\) give the inverse map. Thus \(\Phi\) is the required linear isomorphism. In particular, projectivity of \(T\) implies projectivity of \(D\), because twisting preserves projectivity.
References
The local source declarations below give the construction and proof.
Lean formalization
OAI.KLInvariance.TitsSpace.bottCommon_lift: IntegralPairings.lean:1152.
OAI.KLInvariance.TitsSpace.bottCommon_patch_lift: IntegralPairings.lean:1167.
OAI.KLInvariance.TitsSpace.bottCommon_range_of_patch: IntegralPairings.lean:1182.
OAI.KLInvariance.TitsSpace.bottCommon_range_forward: IntegralPairings.lean:1212.
OAI.KLInvariance.TitsSpace.bottCommon_range_backward: IntegralPairings.lean:1222.
OAI.KLInvariance.projective_of_same_range: IntegralPairings.lean:992.
OAI.KLInvariance.projective_of_coordinates: IntegralPairings.lean:1038.