120V-277VAC are typically rms voltages, so I will assume you want to measure the rms value. All three of your methods should be able to do that with a sufficiently fast ADC.
Method 1 is able to capture the entire waveform without needing a bipolar ADC input, so it is suitable for use with ADCs that can only measure positive voltage. It relies on the midpoint voltage being set accurately, but keeps away from the ends where the ADC may be less accurate.
Method 2 could get up to 1 extra bit of resolution from the ADC if the reference is set for less than 5 V full scale. Whether that will be significant or not depends on the ADC's resolution and noise. An 8 ADC would probably benefit, but a 16 bit ADC might be all noise in the lower bits anyway.
The rectification circuit could introduce error. A simple diode rectifier would be very bad of course, but even a 'precision' rectifier could distort fast changing parts of the waveform.
Method 2 loses the sign, which could be a problem for rms power measurement. On the other hand if the ADC is very slow then it may be the only method that works at all, since it can at least get a reasonable average over a large number of samples (whereas the others would average to zero volts).
Method 3 is like method 1 except that it avoids possible error caused by the level shifting. However it requires a bipolar ADC input to capture the whole waveform. If the ADC input is unipolar then the bottom half of the waveform will be lost, and then accuracy will be reduced if the waveform is not symmetrical.
In most cases method 1 is probably the most suitable. It works with the majority of ADCs, and should be able to get good accuracy when measuring non-sinusoidal waveforms such as are produced by light dimmers etc. It is also suitable for measuring true power (with simultaneous current measurement), and requires very little circuitry to implement (1 resistor!).
