Hi,
to my extend, the color in the tree will just mean if the branches are upon or below the inferred regression. To better inspect which sequences that may be problematic and may need to be removed, I would rather take a look at the `Residuals' panel in TempEst and select the points which are appear far away from the central distribution. Keeping those data points selected in the `Residuals', you can then navigate back to the 'Tree' panel and see which are the names an location of the problematic taxons that may be removed. You can find in this work from Baele et al., a nice tutorial on setting up such TempEst analysis.
Best,