Dear Ryan,
For the mean across channels if you do:
D = spm_eeg_load
m = squeeze(mean(mean(D(D.indchannel({'Cz', 'Pz'}), D.indsample(0.5):D.indsample(0.9), :))));
save m.txt m -ascii
then there will be no warning. For looking at a single channel you can do:
D = spm_eeg_load
m = squeeze(mean(D(D.indchannel('Cz'), D.indsample(0.5):D.indsample(0.9), :), 2));
save m.txt m -ascii
Note that if you start looking at all the electrodes one by one then you should correct for multiple comparison across channels and then if you don't really have any prior hypothesis (unlike what you said before) it does make sense to go back to the SPM approach and compare scalp maps averaged over time, which might be more sensitive than Bonferroni correction across channels.
Best,
Vladimir