It is all very well to complain about the 8051 not being optimized for C but that is a bit like complaining that buggies were not designed with engine mounts.
The 8051 was optimized for very specific, small-scale programming at maximum efficiency and its survival says it did a good job. Keil has indeed done a good job producing efficient code that fits the quirks of the limited internal RAM and four register banks. I suspect few can code much more efficiently in assembly--certainly not with less programmer effort!