Wavelet Phase Spectra vs. Power Spectra

I'm using the continous wavelet transform for analysis of a time-series
which I know contains activity in two known frequency regions. The power
spectra, |Ca,b|�, seems to give no indication of what I suspect to find,
however the phase spectra, tan��(Im(Ca,b)/Re(Ca,b)), very well shows
activity in the correct scales. What is going on? What physical meaning
has a phase spectrum? I'm using the complex morlet wavelet. In classical
Fourier analysis, the phase spectrums shows the phase of the corresponding
frequency components. But I get confused when I think of this with the
wavelets; they are actually translated across the timeseries and therefore
they have all possible phases for each scale? I would really like an
explaination.

Thank you!