Help me with TSShapeAlloc, please?

I'm trying to use it in a way that it's not meant to be used, and maybe I need to try a completely different approach or something, and I'm willing to do that, but I'm at my wits end as to the correct approach.

I have a series of vertices I'd like to insert into a TSShape. I figured the best way, since it has a TSMesh, was to make one of those. But, digging in the code, I find this TSShapeAlloc. I'm sure it has something to do with making serialization / deserialization go faster, but also clumping data together to facilitate cache hits... ok, this is a good thing. But... how do you use it correctly? I thought I understood the documentation and the examples... here's what I have so far, but I'm getting a crash:

[i]// ... In a function named TSShape::readMDL(Stream* s); ...[/i]
            [i]// Assume pVerts[] is a list of vertices, and I copy it into mesh.verts[/i]

            Point3F* pVertAddress = new Point3F[pNode->m_usVertexCount];
            mesh.verts.set(pVertAddress, pNode->m_usVertexCount);
            for(UINT16 x = 0; x < pNode->m_usVertexCount; x++)
            {
               Point3F& vert = mesh.verts[x];
               vert.x = pVerts[x].m_x;
               vert.y = pVerts[x].m_y;
               vert.z = pVerts[x].m_z;
            }

            // Now the nastiness starts.  
            TSShape::alloc.setWrite();
            mesh.disassemble();

            // Get the buffers and their sizes
            S32 bufferSize32 = TSShape::alloc.getBufferSize32();
            S32* buffer32 = new S32[bufferSize32];
            memcpy(buffer32, TSShape::alloc.getBuffer32(), bufferSize32 * 4);

            S32 bufferSize16 = TSShape::alloc.getBufferSize16();
            S16* buffer16 = new S16[bufferSize16];
            memcpy(buffer16, TSShape::alloc.getBuffer16(), bufferSize16 * 2);

            S32 bufferSize8 = TSShape::alloc.getBufferSize8();
            S8* buffer8 = new S8[bufferSize8];
            memcpy(buffer8, TSShape::alloc.getBuffer8(), bufferSize8);

            // Now turn around and do it the other way.
            TSShape::alloc.setRead(buffer32, buffer16, buffer8, true);
            alloc.setSkipMode(true);
            [b]TSMesh::assembleMesh(TSMesh::StandardMeshType, true);   <---- crashes here[/b]
            S32 buffSize = TSShape::alloc.getSize();
            TSShape::alloc.doAlloc();
            mMemoryBlock = TSShape::alloc.getBuffer();

            TSMesh* pMesh = TSMesh::assembleMesh(TSMesh::StandardMeshType, false);
            meshes.set(pMesh, 1);

            delete[] buffer32;
            delete[] buffer16;
            delete[] buffer8;
         }
         break;

At the crash point, the call stack is this:

TSShapeAlloc::get32()
TSMesh::assemble(skip = true)
TSMesh::assembleMesh(meshType = 0, skip = true)

TSShapeAlloc is corrupted at this point and the nice little buffers that I set with setRead() are pointing in lala land.

Any idea what I'm doing wrong? Or better yet... this looks really hackish to me... what would be a better way of doing it?