The basis for each of these derivative specific items ....

Richard,

What this is all about is whether a toolset (in this case Keil) should handle every derivative specific nook and cranny there is.

This raises an interesting question: if Keil had to raise the price by, say, $100 to fully support every nook and cranny of the Dallas derivatives, should I, who do not use these chips, pay the cost of Keil supporting you with the Dallas derivatives?
Just take a peek at the Keil devicelist and you will, if you glance at a few datasheets, see that to support everything, including the "Acme" processors, the simulator would, as a minimum, quadruple in size.
Kail has implemented some support of some derivative specific issues where it counts in the compiler. e.g. some support of double DPTR.

I think - NOTE "I think" - I am not discussing this, just state my opinion, that Keil et al should fully support the standard '51 instruction set only.
If any nooks and crannies of a specific derivative were to be supported, the device manufacturer should be responsible for providing a "tool extension" just like in the CPLD world where the manufacturers provide "fitters" for the universal tools.

Erik