Part of the problem is that the RunScribe stops logging into the data file when the step rate goes below some amount. Rather than getting a data point every 0.7s you might get a gap of many seconds or even minutes. The graphing then tries to join these widely spaced points up. Changing the filtering or switching to a scatter plot will not necessarily make graphs more realistic as the raw data is not being stored.
This post has some details on this problem from a set of intervals.
Adding a couple of data points at the start and end of the resting period made the data more realistic. However, the real fix seems to be to log more data points during walking or resting.