When you search about evaluating RPQs using sparse adjacency matrices you find a paper Diego Arroyuelo et al. [^1]. The algorithm is open-source [^2]. Though it involves hand-written CSR matrix implementation making unfair the comparison with SuiteSparse:GraphBLAS [^3] based algorithms implementing RPQs using adjacency matrices since it's unclear why other algorithms are faster/slower.
It'd nice to implement the algorithm suggested in their paper [^2] using SuiteSparse:GraphBLAS [^3] for fair comparison between the alternatives to differentiate whether the approach is better or the implementation.
[^1] https://arxiv.org/abs/2307.14930
[^2] https://github.com/adriangbrandon/rpq-matrix
[^3] https://github.com/DrTimothyAldenDavis/GraphBLAS