You could use the option -keepthr (and, BTW, drop the index options, as 0 and 1 are their defaults). With this option, the correlation coefficidnts will not be thresholded/clustered (ie they remain unchanged), but for most purposes this is probably OK. If in a subsequent analysis step you do need the clustered correlation coefficients, you could always use the intensity subbrick as a mask.
Tom