Roland is concerned --- 90% of people in the country know someone who is unemployed, and Roland apparently takes this as evidence that unemployment is high. This is a somewhat bogus argument. Sharon raises what is at root a very sensible objection --- because 90% know someone who is unemployed does not necessarily mean the unemployment rate is high. I'm going to guess that the vast majority of the population, almost everyone, personally knows someone over the age of 80 --- this doesn't mean that the majority of the population is over 80, but just that there are enough 80+ year olds distributed in the population densely enough that virtually everyone knows at least one person like this.

Sharon's argument is essentially a probabilistic argument --- if every person knows 50 people, then if 5% = 0.05 = 1/20 of all people are unemployed, then there's a very high chance that of the 50 random folks one person knows, at least one of them is unemployed. One subtle thing about this argument --- this argument works well if 1/20 is a true probability, that is, if we can pick essentially any group of fifty people, and the probability that any one person is unemployed is 1/20, and this probability is more or less constant as we look at different groups of 50. If the probability is not fixed --- if it's higher in some places and lower in other places --- then that changes the nature of the argument.

Suppose, for simplicity, the entire country consists of just ten cities, with small and equal populations. Suppose people know each other well within each city, but essentially no one from one city knows any one from any other city. (This is a highly unrealistic scenario, just to demonstrate the logic.)
SCENARIO #1: 5% of the people in each of the cities are unemployed --- then, most people in each city would know someone unemployed. Because there's an even distribution, about 90% or more of the population would know someone unemployed, even though only 5% of the population is unemployed. This assumes more or less even distribution of the unemployed. This is consistent with Sharon's argument.
SCENARIO #2: Now, consider an extreme of population concentration --- 50% of the people in City #1 are unemployed, and everyone in the other nine cities are fully employed. Here, everyone in City #1 would know someone unemployed, but City #1 is only 10% of the population. Assume the folks in different cities don't know each other, so no one else know the unemployed in City #1. Thus, 5% of the population is unemployed, and only 10% of the population knows someone unemployed.
In other words, if the unemployed are more-or-less evenly distributed in the population, that is, if we could go anywhere, any region or any subgroup of the population, and the probability that any single person is unemployed is about 5%, then we could have only 5% unemployment but a very high percentages, maybe 90%, who know someone who is unemployed: again, this is essentially Sharon's argument. BUT, if the unemployed are not even distributed, and are instead concentrated geographically in particular places, then Sharon's argument would no longer be valid.

That's why (B) is an excellent answer, the best answer for this question.