We present a graph-theoretic framework for constructing floor plans that support non-rectangular modules, with particular emphasis on L-shaped and T-shaped geometries. Unlike traditional approaches that primarily focus on rectangular modules and outer boundary constraints, our method explicitly incorporates structural restrictions that arise when realizing more complex module shapes within rectangular floor-plan representations. The framework is based on triangulated graphs and investigates how algorithmic graph theory techniques can be used to embed L and T-shaped modules while preserving prescribed adjacencies. We show that not every triangulated graph admits such realizations and identify structural limitations that prevent the existence of the desired module geometries. To capture these limitations, we introduce a shape-preservation constraint that ensures module geometries cannot be altered through boundary deformation, as such changes would either increase the combinatorial complexity of neighboring modules or violate adjacency relationships. We propose a linear-time construction algorithm based on a prioritized canonical ordering that realizes L and T-shaped modules in graphs containing at least one internal K4, or two internal K4 subgraphs satisfying specific existence conditions. The algorithm is simple, constructive, and directly implementable, making it suitable for practical floor-plan generation workflows. We conclude by discussing extensions to additional module shapes and broader classes of supporting graph structures.