The Mopar box's dwell is not constant either. It's a very simple design from an era where transistors were expensive and using an integrated circuit in a module would have been unthinkable. (At one point, Chrysler patented a voltage regulator design, which they never used, that had only one transistor because a second one would drive up the cost.) The way it behaves is that it normally keeps the transistor in an on state. When it receives a pulse from the distributor sensor, this shuts off the transistor (like opening the points) for a duration controlled by a timer circuit. The circuit that detects the pulse has a very short response time - but it's long enough that it retards the timing a tiny bit. Since it's a somewhat fixed amount of time that delays the spark, but there's less overall time in between sparks, you get a measurable amount of timing retard at high RPM.
Only real ways around this are to make a more sensitive trigger circuit, which would have meant more complexity and cost, and possibly a risk of sparking in response to electrical noise. Modern computer controlled ignitions either have a variable that predicts the amount of delay, or have it factored into a computerized curve that's a bit more complicated than anything you can do with a mechanical advance.
The race modules have a shorter timer circuit - if the off time is too long, you don't get enough dwell to get decent spark energy at high RPM. Conversely, a shorter off time and lower resistance ballast resistor makes for more heat at low RPM. I haven't personally investigated Chrysler's report that their ignition combinations tailored for high RPM overheat at prolonged operation below 2000 RPM, but it's consistent with the way they operate.
