In parts C and D you were expected to actually count! How many of you simply used the normal approximation! You wanted to check to see how well this particular data set actually approximates the normal case! (and it does very well...).
On part E., it was good to actually cite Chebysheff's rule - but it was important to note that the results were not particularly close, because the rule is so conservative.
As far as part G, you should note that this is not a random sample of cities (it is a sample of "well-known" cities), and hence should not be considered representative of cities in general. For example, these well-known ones will tend to be large. It is interesting to note that some cities are so large that they "make their own weather" (I've heard for example that the concrete and asphalt surfaces of Atlanta lead to local changes in the weather).
The most important problem in the whole exam, perhaps, was number 3 - unfortunately not a single person got it right! The important thing to do was to ask whether a normal approximation was appropriate - and it's not. Only a few students questioned the normal approximation - most computed using the Z-table blindly! If you did, then you've been lying with statistics....
We know this because distances have a mean of 100, a standard deviation of 80, and must be positive - hence we can't possibly get a lovely bell-shaped curve out of that. We can go at most 1.25 standard deviations to the left of the mean.
Chebysheff's theorem can be invoked to tell us that we can have at most 25% of the distances greater than 260 (since this represents 2 standard deviations from the mean, and Chebysheff's theorem states that at least 75% of the data must be found within two standard deviations from the mean.
Furthermore, if we had all data at exactly one standard deviation from the mean (hence, an equal number of data values of 20 and 180, then we would have a standard deviation of 80 and a mean of 100 - hence it's possible that 0% of the data are at a value of 260 or greater.
The histogram has got to be skewed, since the mean and the median differ. The median is the value that separates the data into two equal halves to the left and right. The mean is the value at which the histogram would balance (the mean being the fulcrum of the teeter-totter).
As far as extreme values, 0 or less is unacceptible, since the time must be positive; as for the greatest, going out 3 standard deviations or so makes a lot of sense. Both of the examples above got out to 34 for the largest value - good job!
Some people had all the data within one standard deviation - this is impossible, unless all data are evenly split between one standard deviation to the left and one to the right.