In case this was somehow lost in the data system, here is my first answer yesterday.

Some of this probably depends on your intentions and level of interest.

As you already have found, stream sediment varies with flow, and the hysterisis effect due to differences in energy within the system occurs during rising and falling stages of the hydrograph.  There are a variety of reasons for this, such as the rising stage has the added energy of rainfall detachment, and access to any fresh soil disturbance by natural or activity related.  As you may have learned, although difficult within one month, besides the rainfall intensity differences in storms, there can be seasonal differences due to things like leaf fall, activity of organisms, effect of temperature affecting water density, etc.  I am sure with a literature search, you will find some good references.

You may not be aware of Dr. David Rosgen, but he has a book as well as available papers on his website (www.wildlandhydrology.com) that might be interesting that presents his approach to discussing and dealing with the stream power differences that affect bedload and suspended sediment loading for individual streams.  His approach is intensive and typically well beyond what might be found in a few months, with expectations of obtaining a bankfull flow measure with both bedload and suspended sediment concentration to be used as a normalizing factor in adjusting other data taken at other flow levels to develop a curve of normalized flow and normalized sediment.  

However, I am not sure the degree that his approach considers the hysteresis effect, and that might have to be a manual separation of curves, rising stage and falling stage.  Some stable stream types and watershed conditions are apt to have better ability to compare storm to storm data.  Watersheds with much variety of activity, unstable stream types or conditions would likely to have more variability.  

I hope in collecting this sediment data, you are able to have corresponding rainfall intensity and water level - stream flow - stream power  and watershed as well as stream type data to help evaluate the differences associated with land erosion and sediment delivery, and stream bank and channel erosion and sediment delivery.  

Within the specific hydrograph and perhaps this is really your question, when to start and when to stop.  You may note with some storms a flush of material during the rising stage, and if you have not considered, you might want to analyze to separate out the organic sediments from mineral types, and also analyze particle sizes present, recognizing that this adds more work and may not end up being that useful for your intentions.  These flushing storms often occur after significant periods of little or no rainfall and channel flushing, such as after the growing season where flow declines to its annual minimums.  This first flushing storm may show up as different from others as fines are being entrained and transported as these events tend to be more severe to produce much of a hydrograph in the dry soil conditions.  These are just generalizations.

But assuming you have full detail of sediment with stream stage or flow, I would include all the readings from when the stage increases, probably separate at the peak, and then all the readings in the falling stages of the hydrograph to where sediment readings level off with very little or no change with additional time. Another idea is using the inflection point where the hydrograph changes slope from a fall slope to a baselevel flow slope.  I dont know if there is any exact standard for separation of these type data.  You might also consider how long it takes to return sediment concentration to pre-storm data. Again, some of these type decisions may depend on your intent.   

In case this was somehow lost in the data system, here is my first answer yesterday.

Some of this probably depends on your intentions and level of interest.

As you already have found, stream sediment varies with flow, and the hysterisis effect due to differences in energy within the system occurs during rising and falling stages of the hydrograph.  There are a variety of reasons for this, such as the rising stage has the added energy of rainfall detachment, and access to any fresh soil disturbance by natural or activity related.  As you may have learned, although difficult within one month, besides the rainfall intensity differences in storms, there can be seasonal differences due to things like leaf fall, activity of organisms, effect of temperature affecting water density, etc.  I am sure with a literature search, you will find some good references.

You may not be aware of Dr. David Rosgen, but he has a book as well as available papers on his website (www.wildlandhydrology.com) that might be interesting that presents his approach to discussing and dealing with the stream power differences that affect bedload and suspended sediment loading for individual streams.  His approach is intensive and typically well beyond what might be found in a few months, with expectations of obtaining a bankfull flow measure with both bedload and suspended sediment concentration to be used as a normalizing factor in adjusting other data taken at other flow levels to develop a curve of normalized flow and normalized sediment.  

However, I am not sure the degree that his approach considers the hysteresis effect, and that might have to be a manual separation of curves, rising stage and falling stage.  Some stable stream types and watershed conditions are apt to have better ability to compare storm to storm data.  Watersheds with much variety of activity, unstable stream types or conditions would likely to have more variability.  

I hope in collecting this sediment data, you are able to have corresponding rainfall intensity and water level - stream flow - stream power  and watershed as well as stream type data to help evaluate the differences associated with land erosion and sediment delivery, and stream bank and channel erosion and sediment delivery.  

Within the specific hydrograph and perhaps this is really your question, when to start and when to stop.  You may note with some storms a flush of material during the rising stage, and if you have not considered, you might want to analyze to separate out the organic sediments from mineral types, and also analyze particle sizes present, recognizing that this adds more work and may not end up being that useful for your intentions.  These flushing storms often occur after significant periods of little or no rainfall and channel flushing, such as after the growing season where flow declines to its annual minimums.  This first flushing storm may show up as different from others as fines are being entrained and transported as these events tend to be more severe to produce much of a hydrograph in the dry soil conditions.  These are just generalizations.

But assuming you have full detail of sediment with stream stage or flow, I would include all the readings from when the stage increases, probably separate at the peak, and then all the readings in the falling stages of the hydrograph to where sediment readings level off with very little or no change with additional time. Another idea is using the inflection point where the hydrograph changes slope from a fall slope to a baselevel flow slope.  I dont know if there is any exact standard for separation of these type data.  You might also consider how long it takes to return sediment concentration to pre-storm data. Again, some of these type decisions may depend on your intent.   

As Dr Hansen clearly said, I will focus on the changes slope on the hydrograph. These changes are directly associated to those of the precipitation and currents, and so they relate to the physical factors important for you. Just take care that the maximum of solid discharge sometimes occured just after the one of liquid discharge.