To perform the average correctely, I believe all the samples should have the same windows time. In your case, Brainstorm uses the first subject to define the time window and then applies it to all your subjects.
Thanks!! Finally, I averaged the interval of each subject. In this way, I got one file of 1 millisecond by subject. After that, I click on averages files.
Another solution can be to select all the trials in Process1 and call the process "Standardize > Uniform epoch time", and then average the files.
This would work as expected if all the trials have exactly the same number of time samples.