Yaron Hollander is an experienced transport modeller, previously at Transport for London and now an independent consultant. He has written and published a book for beginners, which I warmly recommend. The style is accessible, jokey even, with cute hand-drawn illustrations. I learned a lot both from the the main expository text, and also from Part II, which discusses critically the culture of transport modelling.
Yaron argues that models work best for short term predictions of small scale interventions, such as fine-tuning traffic signal timings. But in practice models are mostly used to support investments in long lived infrastructure: here the modellers may understand the shortcomings, which their clients may prefer not to recognise provided the model outputs justify the planned investment.
Highways England has a major development of modelling underway to justify intended investment in 100 inter-urban road schemes. This ambitious project involves using mobile phone data to identify origins and destinations of trips, and the application of of the very fast computation that makes computer games possible. However, I would be surprised if many of the model outputs failed to confirm the investment decisions, given the scope for judgement about model inputs reflecting future developments, particularly traffic growth.
My conclusion is that transport modelling as it has developed is decidedly problematic. Models generally serve to confirm prior expectations, rarely to illuminate decisions.