Pairwise Spearman Correlation for Independent Components vs EEG Channels recordings

When estimating the pairwaise correlation co-efficients at the original channel level recordings, I observe that they are less than those of calculated at the components obtained between Independent componets which are obtained by Independent Component Analysis (ICA). Is this possible to have?

While the spearnman correlation co-efficient can be calculated to estimate for linear independence, should it not be less when it is estimated at the components level. This is also because of the reason that the linear independence is a weaker form of statistical independence (which ICA trying to achieve),

Appreciate any help to understand.