The proportionality between differential cross sections at vanishing linear momentum transfer and Gamow-Teller transition strength, expressed in terms of the textit{unit cross section} ($hat{sigma}_{GT}$) was studied as a function of target mass number for ($t$,$^{3}$He) and ($^{3}$He,$t$) reactions at 115 $A$MeV and 140 $A$MeV, respectively. Existing ($^{3}$He,$t$) and ($t$,$^{3}$He) data on targets with mass number $12leq Aleq 120$ were complemented with new and reevaluated ($t$,$^{3}$He) data on proton, deuteron, $^{6}$Li and $^{12}$C targets. It was found that in spite of the small difference in beam energies between the two probes, the unit cross sections have a nearly identical and simple dependence on target mass number $A$, for $Ageq 12$: $hat{sigma}_{GT}=109/A^{0.65}$. The factorization of the unit cross sections in terms of a kinematical factor, a distortion factor and the strength of the effective spin-isospin transfer nucleus-nucleus interaction was investigated. Simple phenomenological functions depending on mass number $A$ were extracted for the latter two. By comparison with plane and distorted-wave Born approximation calculations, it was found that the use of a short-range approximation for knock-on exchange contributions to the transition amplitude results in overestimated cross sections for reactions involving the composite ($^{3}$He,$t$) and ($t$,$^{3}$He) probes.