Fair point, that's true in the direct lineage sense, but, 6809 to 68000 is a similar jump, there is nothing to say the one was based on the other except for general ideas and some addressing modes that turned out to be handy (when writing compilers, rather than writing assembly). Every widening of the databus caused a redesign from the ground up, even if some of the concepts survived. The 4004 was early enough that there was not much installed base to worry about so a clean start for a new chip made very good sense.
But in the 65XX family there is the 65816, a chip that tried really hard to maintain as much backward compatibility as possible. It saw some commercial deployment (Apple, Nintendo). At that point in time backwards compatibility began to have real value and intel really made some lucky calls: the weird addressing modes resulting from the lack of register width eventually culminated in a setup that worked very well for CPUs that were running multi-tasking OS's. The 386 was a very nice match for such code and this model was a major factor in the success of the line (which really was creaking badly with the 80286 out vs the 68K, which effectively had a 32 bit flat model built in because of its ability to run position independent code).
But in 1987, when the 80386 hit GA it was pretty much game over for the rest even if it took a while for the other empires to crumble, only ARM survived and that is mostly because Acorn had a completely different idea about power consumption and use of silicon than Intel did. The current crop of x86 hardware is insane in terms of power consumption and transistor count, ARM is so much more elegant (in spite of its warts).
