A dipole can be physically a half wave, or electrically a half wave.
A dipole that's physically a half wave has a nonzero reactance at the feedpoint. This means it is not resonant, and that the reflections within the antenna are not precisely in phase with the signal. (These are two ways of saying the same thing.)
The electrical length is determined by the time taken for the wave to traverse the antenna. Making the dipole a little shorter until the reactance is zero makes the dipole electrically a half wave. Now it is resonant.
Making the antenna resonant doesn't make it better per se, but since any low-loss feedline (like, any one you'd use in practice) has a negligible reactance to its characteristic impedance, an antenna that is not resonant (which means a non-zero reactance at the feedpoint) must necessarily be mismatched to some extent. A mismatch means additional SWR loss.
An antenna that is resonant is not necessarily matched to the feedline, or even at minimum SWR, unless it just so happens the feedpoint impedance and feedline's characteristic impedance happen to be the same. Resonant dipoles typically have an impedance around 75 ohms; bending the dipole into a "V" decreases the feedpoint impedance and it's usually possible to get it down to 50 ohms. So it's usually possible to match a dipole to 50 or 75 ohm coax without any additional matching devices, though you'd still want a balun in there.
Of course you could always match the antenna some other way, but most ways you might do it (like adding discrete capacitors or inductors) would introduce additional loss, usually more than is saved by reducing the mismatch losses.
In practice on HF, all this is more theoretical than practical because the feedline losses, and thus the mismatch losses, are low. Any SWR under 1.6:1 is within the specifications of most transmitters and "plenty good enough" for most HF applications. The advantage to twiddling with the dipole shape to get closer to 50 ohms is really just that you then have a wider band where you're under 1.6:1.