Out of phase peaks may appear in a 2D NOESY spectrum due to several reasons:

1. Incorrect phasing: If the zeroth and first order phase corrections are not correct, you can try alternate methods of phase correction.

2. Incomplete solvent suppression: If the solvent peaks are not properly suppressed, the remaining signals can have components that resemble the dispersive component. Try alternate methods of solvent suppression.

3. If the spectrum has been properly processed and there is no problem with solvent suppression, you may still observe dispersive peaks due to zero-quantum coherence. Zero-quantum coherence is generally not a problem for NOESY of large molecules, and many standardized pulse-sequences do not attempt to suppress the contribution of zero-quantum coherence. Zero-quantum coherence may be suppressed by use of z-filters or by using a short (~2ms) spin-lock pulse or a synchronous spin-lock and pulsed field gradient. However, you should be careful regarding the last, as it has the potential for causing damage to the probe - so consult your NMR administrator before trying to modify the standard version of your NOESY pulse sequence.