A couple of dB isn't noticeable by the average listener as most can't discern less than 3dB. I do audible noise testing and analysis on motors and related mechanisms and notice more than most but I somehow doubt I discern anything below 3dB myself. However, flattening response by one dB with better matched components, another dB with a Zobel network, a third dB with an attenuation network, yada yada ... will eventually add up to a noticeable difference.
The results my Minimus 77 rebuild is testament to some of the general rules of thumb. The sould quality is greatly improved. I wish I had some "before" data but a better crossover flattened a decent hump at 2.3KHz. It resulted from not crossing over more than an octave above the tweeter's resonant frequency. I doubt my attenuation network would've hit the right frequency if I had mounted its inductor too close to the speaker or the metal cabinet or the low pass network inductor. It could've easily made things worse.