Book by Timothy Claridge "High-Resolution NMR Techniques in Organic Chemistry" recommends calibrating homonuclear decoupling pulse power using Bloch-Siegert shift. It's somewhere in the "calibrations" section of the book.
Is there a particular reason that it would be inaccurate to set decoupler directly on resonance and calibrate width of low power 90 degree pulse by finding 360 pulse zero crossing?
The method of choice for this calibration would depend on the information that you want and the experiment that you are doing. Are you wanting the 90 flip angle calibration for a low power selective pulse experiment? Or the decoupling field strength in an actual homonuclear decoupling experiment?
If you want a pulse calibration, you can certainly just set the transmitter directly on resonance and calibrate the 90 degree pulse in a normal manner. This is best done on a singlet as precession of off-resonance lines in a multiplet can be a problem at very long pulse lengths (10's of ms).
For calibrating a homonuclear decoupler field level (expressed as gamma B2/2Pi in Hz), yes you can use the Bloch-Siegert shift - and this may indeed be the method of choice. Calculating the homonuclear decoupling field level from the 90 degree pulse calibration is also possible, but requires some additional work. This is because on FT machines homonuclear decoupling is not continuous, but is pulsed in a time shared fashion with the dwell clock. You have to know the effective duty cycle (typically ~10%) in order to calibrate the field strength. I haven't really looked into it, but this might be even a bit more complicated in new systems with oversampling, DQD, direct digitization of the IF, etc. On some older systems (e.g Bruker AMX), the homonuclear decoupling signal could be "injected" into the signal path using a directional coupler. You then have to include the directional coupler loss into the calculation.
The good news is that on modern systems the transmitters are linear enough and the attenuators accurate enough that once you have done this calibration at one power level, it is easy to calculate the field strength at any other power level.
Hope this helps!