Besides the fact that we do not wish to be reliant on human-simplified models of training, there are a number of reasons. For example, what TRIMP calculations could you use? (Running with mobile phones, the majority of runners do not record heart rate, and then those people that do record it are more and more using optical sensors where reliability is often questioned.)
Then there is the bigger issue that TRIMP tells you nothing of the type of training your should do next, and ubiquitously TRIMP just does not work to gauge effort. This is not only true when trying to compare efforts between modalities (e.g. running and cycling), but also within a single discipline. As an example, TRIMP has been demonstrated to out by a factor of 3 when comparing the recovery time ‘required’ following endurance vs sprint workouts; and a factor of 6 between different activity types.
An easy way to look at the issue is that a 30 minute running speed session may have a very similar TRIMP score to a 90 minute easy run. However, they will have very different physiological effects and recovery needs – yet TRIMP states no difference.
Furthermore the standard models that utilise TRIMP (namely Training Stress Balance) have not been scientifically validated.