Small boundary mass flux of scalar leads to null values of concentrations in CS v6.0 and v7.0

Putting a small mass flux Qm of the order of microgram/m2/s ( rcodcl (ifac, isca(iscal), 3) = -Qm in cs_user_boundary_conditions.f90 ) leads to zero mass fraction values, when larger values of Qm give finite values. I tried different turbulence models (k-eps, k-omega-sst) with different wall functions (2 scales + scalable with/without log laws). My y+ are quite small (<1).
Following one advice from Y.Fournier, I rewrote the routine as a volumic source term close to the emiting surface, but the problem persists.