Deep Breathing
As TZ750 tuners discovered when they tried to comply with the infamous AMA Tech Bulletin 78-1, it is conditions downstream that are most important in dealing with a restriction (TB 78-1 was the AMA's restrictor rule, origin of many a seized cylinder as tuners struggled with unfamiliar conditions).
Pull out the Superflow Corporation's little handbook and read their estimate that 40% of the flow resistance across a four-stroke engine's poppet intake exists as diffuser loss - DOWNSTREAM OF THE VALVE SEAT. All that clever porting and shaping is work done to optimise the 60% of the problem that is often all that gets thought about by the tuner.
Let's examine what happens in a convergent-divergent nozzle. As flowing air enters the convergent portion, it must accelerate. As it does so, some of its pressure is lost in exchange for increased velocity. This is a reasonably efficient process because it creates very little random motion; the flow is accelerating, so there are no problems with boundary layers getting thick and separating, creation of horrible eddies, and so on. It is an efficient process.
At the throat of the venturi, the process reverses, and velocity is to be exchanged for pressure. This is just what we want of the air entering an engine cylinder; it has low pressure as it whizzes under the valve, but must thereafter slow down, creating pressure in the cylinder. The trouble is that this process -- called diffusion -- is harder to keep efficient. If the duct cross-section is enlarged too rapidly, in an attempt to slow the flow rapidly, the flow separates from the walls and its energy, instead of creating pressure again, creates eddies and, ultimately, heat. This is just what happened with the first-try restrictors back in 1978; high velocity through the 22 mm restrictor was not efficiently converted back into pressure downstream of the restrictor because the diffuser section was too short and too steep to keep the flow attached. Yamaha crushed the opposition by developing a correct design which used minimum length for acceleration (the efficient, trouble-free process) and devoted all the remaining available length between carb and reed to a narrow-angle, efficient diffuser. Kenny Roberts used a bike equipped with these efficient restrictors to set new race records and lap the field.
Now back to the four-stroke case. At first it appears that we can do nothing about the "diffuser" -- the region outside the valve seat in which the flow dumps out suddenly into the cylinder -- with plenty of flow loss. However, just because we can't design in a classical diffuser doesn't mean there is nothing that can be done. How about half a diffuser? If we can attach part of the flow to a wall or walls, we can get partial diffuser action and increase the efficiency with which port outflow is converted to pressure in the engine cylinder.
In general, tuners are at pains to avoid aiming the intake flow right at the exhaust valves because of the potential for short-circuit loss during overlap. Yet perhaps there are ways to attach the flow to head and cylinder wall that avoid this. Maybe there is a reason why the intake ports of Cosworth designs are high, but not extremely high. Maybe the desirable thing is not so much to get the flow nearly parallel with the valvestem as it is to aim it to attach to the cylinder head - not shoot out into the cylinder as a free and inefficient jet.
If you have a flow bench, try this process of attaching the flow to a surface, and notice the substantial flow gains that attend it. I knew there was a reason not to watch television too much.