Think of it as a frequency/likelihood thing- For the sake of example, lets go with the premise of a TDA system that could reduce the likelihood of an unintended discharge xx%, but at a cost of greater training and probably turnover due to some employees who can qualify on a SFA weapon but not TDA. UDs with injuries are huge liability, but very low frequency. The costs for training and retention/recruitment (plus possible litigation on the part of those borderline employees requesting "reasonable accommodation" via SFA) are pretty significant AND will be moderate to high frequency.
EDIT- There is little consensus as to exactly how much a TDA pistol reduces unintended discharges. While intuitively I believe they provide some extra people management margin for error (see LASD migration from B92 to SFA), I suspect most startle responses are strong enough to pass clean through most all DA triggers. Unless you are proposing 1895 Nagant revolvers that is....