Hi, Jenna-
Sorry for missing this the first time through.
Setting motion thresholds---balancing the removal problematic time points with still having a useful amount of data---is tricky. Indeed, in a number of populations, some thresholds are set higher. It will still be good to verify that motion artifacts are not overwhelming effects on correlation---hopefully reviewing the APQC HTML will help with that.
Re. correlation matrix: are you using 3dNetCorr? When you analyze the errts, then mean of that time series is 0, and within a subject those 0s match across a whole volume. The correlation values should not be strongly affected by that.
A 0 in a correlation matrix would mean that the average time series for all voxels in an ROI are identically 0. That shouldn't really happen, if the ROI is within the brain mask.
--pt