Be careful with the terminology. And yes, just because the wrong terminology appears in a few papers that are improperly peer reviewed, does not mean this terminology is correct. Bioconcentration factors are "generally" measured between organisms (i.e., animals) and the water in which they live - and they assume that steady-state has been reached. For animals, they also assume that the dominate route of uptake is through the water phase, and not through ingestion. If you attempt to normalize your "ratios" to the soil (micrograms per kg of soil), it will have no thermodynamic basis, and metal adsorption to different soils can be vastly different. You CAN calculate "distribution coefficients", which is not semi-empirical "thermodynamic" values. If you do this, yes, you should base the ratio on the metal remaining in the soil, not the initial concentration (you want the steady-state ratio). Better if you use the water phase concentration in the soil - if you can grow the plants in the presence of a saturated soil. I am not a "plant" person, so you should refer to some highly cited papers to see how to best process the data.

The mass of soil is typically so great relative to the mass of plants that you cannot readily measure a change in the contaminant concentration in the soil.

I agree with the first answer. There are many considerations about the parameters in the soil that can affect the distribution of your analyte. Translocation factors are other distribution quotient often used in phytoremediation studies that only involves the plants not the soil. In any case, here's a reference that can be useful:

Zhang, Y., Wang, X., & Ji, H. (2020). Co-remediation of Pb contaminated soils by heat modified sawdust and festuca arundinacea. Scientific reports, 10(1), 1-13.