If you conduct soil organic matter tests, you will find that each species of grass will sequester a particular constant in the soil, and generally annuals produce the least, and the bigger the perennial the more soil carbon is produced, but there may be a limit with bamboo in that instead of sequestering it in the soil, it stores the carbon in the stems.
You will find with the perennials especially the bunchgrasses in arid areas, like desert parts of the planet, you can have populations of old plants and low carbon around the plants. That is because grazing animals have removed the carbon, to below the native grass seedling survival levels. If you use reproducing populations where you see young seedlings among the older plants, that will give you a more accurate soil carbon constant for that species. You will find that for each individual species, the soil carbon percentage is very constant.
Aboveground carbon in the leaves and stems of grasses should not be counted as "sequestered" because that is easily removed by grazing animals for example, only count the soil carbon. By checking the soil carbon with radiocarbon dating, you may be surprised how long the soil carbon is removed from the carbon cycle, hundreds to thousands of years.
Henderson et al. Greenhouse gas mitigation potential of the world's grazing lands: Modeling soil carbon and nitrogen fluxes of mitigation practices. Agriculture, Ecosystems & Environment, Volume 207, 1 September 2015, Pages 91–100
I have also found these two articles that focus in analogous themes where maybe you can find some useful information in the discussions (especially the first article, Fig. 8) or in the references, for example:
Connant et al. Grassland management and conversion into grassland: Effects on soil carbon. Ecological Applications, Vol. 11, No. 2 (Apr., 2001), pp. 343-355
O’Halloran et al. (2013) Regional Contingencies in the Relationship between Aboveground Biomass and Litter in the World’s Grasslands. PLoS ONE 8(2): e54988. doi:10.1371/journal.pone.0054988
Another option would be to look for this information separately for each ecosystem of your interest.