A nearly straight tube can now be modelled well enough to match real playing behaviour. Add even one more variable, such as a widening cone, or several, as in brass instruments, and the problem becomes far harder. Shape complexity, not the instrument's familiarity, sets how far a model can reach.
An instrument is not a self-contained sound machine. The vibration is driven and kept going by the performer, so any account of how an instrument works is incomplete without the person playing it. Where the listener stands matters too, because instruments send different air motions in different directions.
For a long time the science of wind instruments could only say which notes a tube of a given shape would resonate at. A real understanding has to model the whole instrument at work, including the uneven, non-proportional push that starts and sustains the vibration. Only then do many of an instrument's musical qualities show up in the model and match experiment.