The problem with the options for LCModel (and the same is true for Tarquin) is that it assumes a homogeneous voxel of one tissue. Once you have different tissue components in the voxel then it is not possible to write the signal as the linear product of concentration and attenuation. My preference is to calculate the combined scaling factor as described in Near et al. 2020 and then pass it in as one of the two parameters (WCONC OR ATTH2O), setting the other to 1. This means that the results returned from LCModel will be in mM which is convenient (e.g. if you just want to share the output pdf etc.). This of course does not work if you wish to use metabolite specific T2 values in which case you have to make the corrections after the LCModel step as Georg suggests. However if applying a global value I find it easier to calculate it first and supply it to the quantification program.
Suspect has a function in its fitting module which will calculate the molar scaling factor you need to supply. It uses the form from the Near paper along with tissue parameters from Gussew 2012, although these can be selectively overridden if desired. If calling Tarquin from Suspect you can pass this parameter directly as aq_factor and it will set the WCONC and ATTH2O params for you, but this is not yet implemented for the LCModel integration.