Is the 95% Ci of the mean wider or narrower than the range of the data?
It depends on sample size. This graph shows two data sets, with error bars showing the 95% confidence interval (CI) of the mean.
The Ci of a mean is computed as the sample mean plus or minus the SEM times a critical value from the t distribution which depends on sample size and degree of confidence you want (usually 95%). With large n and 95% confidence, this critical value equals 1.96. With smaller n, the critical value is larger. With n=2, the critical value is 12.7 and the CI extends beyond the range of the data.
This makes sense. The goal of the CI is to bracket the true population mean. With n=2, it is not all that unlikely that both points are higher, or both points are lower, than the population mean. So the Ci must extend beyond the range. With large n, the population mean is likely to be pretty close to the sample mean, so the CI is narrow -- much narrower than the range of the data.