Analog components have various nonlinearities that account for their signature contribution, subjectively negative or positive. Designers select from those components for desired effect and based on economics, availability, and personal biases.
Digital components add other considerations like sample rate, bit depth, jitter, latency, and perhaps most importantly, algorithmically - the software. Various codec filter algorithms have audible impacts, and when an encode using one is decided using another, the effects are quite unpredictable. Extended over the entire 'chain' from studio to master to mix to distribution, and finally to your listening room, attributing audible changes to a single item requires extraordinary diligence to eliminate confounding variables, confirmation bias, uncontrolled variables and a bunch more. 'Accuracy' is euphonia. One likes the sound, it agrees with the preconception of what something should sound like and, ipso facto, it becomes correct. Measurements, simplistic by their very nature, provide only the vaguest validation. What measurement, short of tweaking the treble tone control, accounts for the oft-used 'lifting a veil' expression of newfound system clarity. And how many veils are layered over my system in the first place?
It's all good fun, and fascinating, in a certain way.