I think your first point has really encapsulated the core issue here - namely that differentiation between fastest hardware and highest benchmark score. What these 3D benchmarks were originally designed to do and what they have come to mean are two distinct and separate things. Basically the meaning of the benchmark is derived. If you as a gamer purely want to test your system and see how it compares against other similar system then all is good. That was after all the original purpose of the Graphic benchmarks was it not? But 3D benchmarks and have evolved meaning into something different than the original intention. They [3D benchmarking] are now a competitive sport. (Let us not quibble about the meaning of the word sport I am using it in a generic sense.) And as such the same forces are at play as found in any sport. "Citius, Altius, Fortius" (faster, higher, stronger). Those sentiments apply to benching as well as to any athletic pursuit I would suspect. And what is the gauge that is used to measure your prowess in benching? Why synthetic benchmark scores of course.
Moving on to the second point you made, software. That is a difficult one to determine just where is the line drawn. Do we then insist that only a certain motherboard bios be used by all who bench with that motherboard? Do we take it even further and insist that a particular version of an OS be used. After all we want the playing field to be even and only test hardware...or do we? Just as a software package in a Formula 1 bike or car, can improve the efficiency of a hardware package so do software packages improve the performance of hardware in PC's. The notion that hardware is independent of the software in not completely accurate. They co-exist - in fact to use a trendy term their relationship is in fact symbiotic and the product a synthesis of all the parts. Differentiation between where hardware ends and where software begins is becoming (in the component design world anyhow) very difficult. I suggest to you that software does in fact positively on how the hardware is better able to do its work and as such has the same value as does hardware modification or tweaking. Legally of course.
.
Your last point has merit and at times I can fully agree with your view - but it is a particular position that you are taking namely what I call the hardware stance. (Apologies to Danial Dennett for messing with his intentional stance notion.) That hardware centric stance places more value on hardware and the results achieved with hardware. Software, bios or driver modifications and testing are just as valuable (in them selfs not only as a outcome produced) as those hardware adjustments and modifications. It is surly the synthesis of all the various elements that makes a great over-clocker and bencher. Access to hardware, the willingness to put in the hours of testing in (both hardware and software testing) and of course having the courage to actually attempt what no-one has done before - these are all needed to be a record breaker.
