Its always difficult to comment on an image alone, especially if it doesn't represent the entire captured data.
However, there is likely to be a plausible reason as to why you got the "error" in the measurement on the occasion that your image represents. For instance, if there was also a negative going spike somewhere else in the capture period then that could have influenced the measurement, because the 'pulse width low' measurement takes into account anything that could be considered a negative going pulse (which should be what you might 'reasonably' expect to be a pulse) over the entire capture period. If there is more than one pulse with different widths, the software can't know which you are intending to measure, so it gives you an average of the pulses, as mentioned in the snip from the context sensitive User Guide, below:
The following Gif file also demonstrates how this works in practice when measuring the whole trace or just between rulers:
(Just to pre-empt any misconceptions, the measurement statistics, such as Min and Max refer to the pulse measurement over multiple captures rather than the measurement of multiple pulses during a single capture)
So, if you can't remove the source of the error pulse when capturing the data then, for the pulse that you want to measure, you can just isolate its position, in the captured data, using the rulers (as I was showing in the Gif file above) so that the measurement automatically defaults just to the location of your required pulse.
If, on the other hand, this doesn't answer your query then could you post the data file, so that we can get a complete understanding of what is going on.