If I knock up something in excel, and other people use that thinking it is how things are, but do not refer to the reality the model represents because there is no way to do that, how useful is the model?
Have you ever looked at old medieval beastiaries? Information is being relayed, but how useful is it?
https://alexanderadamsart.wordpress.com/2019/07/30/the-besti...
Without a means to view the underlying thing the model is meant to represent, to check and correct one's misunderstandings, how useful can the model be?
you are mixing up models and simulations. A simulation can be a great way of exploring something unknown if you understand the rules in enough granularity. Models can be judged by their ability to make accurate predictions. Electrical engineering depends on both a great deal, I expect you agree that it is useful as a field given the method of communication we're using.
For chemotaxis, while they can't observe the underlying processes directly, they result in phenomena that are observable, which can be compared with the same phenomena predicted by simulations and models (the same is true of all scientific fields).