TMM stated the case rather well. I made a graphic to illustrate this concept (see attached pdf file). Imagine a small driver mounted in various baffles. The sound wave radiates from the driver along the baffle. I illustrate 4 different frequencies (wavelengths) and super-impose these on various baffle shapes.
Starting with a driver centrally mounted in a square baffle (near worst case), the transition from 2-pi to 4-pi radiation happens over a very short frequency range. Even worse is a driver centrally mounted in a circular baffle... here the transition happens at abruptly at a single frequency.
Moving to a rectangular baffle, things improve quite a bit, and now the transition from 2-pi to 4-pi takes place more gradually. Even better is a rectangular baffle with the driver offset horizontally. The final graphic is a trapezoid baffle, and clearly this results in a very smooth transition, but is it really better than the rectangle with a horizontal offset? Maybe not. It is possible to select an optimal rectangular baffle with an optimal driver placement which will result in a very smooth transition, one that is hard to improve upon even with a trapezoid shape.
Considering the extra complexity and cost of making a trapezoid cabinet, it may not be worth the effort.
Of course, if you are in the business of selling $20,000 - $100,000+ speaker systems (like Avalon), any possible improvement is worth the effort. Your customer base expects you to "spare no expense" in the pursuit of the finest possible sound reproduction. Even if the a design feature is very expensive, if it has even the slightest possibility of improving the sound, your customer expects you to incorporate that design feature.
