You cannot compare the memory used by Windows 95 with all the other versions you listed. It was a completely different code base and architecture and, as such, any such comparison is meaningless.

You're comparing a single-processor single-user operating system running on x86 real-mode with a multi-processor multi-user operating system with 32 and 64 bit capabilities and virtual memory, process isolation, etc.

It's akin to comparing a bicycle to a car and faulting the car for having twice as many wheels and therefore "wasteful" while disregarding all the additional functionality.

So your very first data point is incorrect and skews the entire comparison. You are correct, of course, that Vista required more memory than it should have and that was indeed one of the problems but it was definitely not the main reason the release failed.