Let's assume there are no faults with setting up the instrument or parameter voltages and the like. Let's also assume the assay worked; that is, control cells looked relatively live, and treated cells looked nicely dead. The possible problem I've been coyly hinting at therefore, is in the analysis.
What's typically the first thing someone does when they analyze their flow data? They make a forward scatter (FS) versus side scatter (SS) plot and gate on the 'live' cells. Obviously, if you're trying to determine frequencies of live and dead cells, you wouldn't use a 'live' gate, you'd extend the gate lower on the FS parameter to include the dead cells. However, when you do so, you could be including small debris and bits of cells, and it's not exactly fair to count fragments of cells as events. So, herein lies the dilemma: You'd like to count whole cells, while drawing a light scatter gate that encompasses both live and dead/dying cells. Hence, My 3-step approach to gating Annexin V data appropriately.
Step 1: Display your Annexin V channel versus your Membrane Permeability Channel (for ease of typing, I'll assume we're using Annexin V FITC and PI) on the entire, ungated data set, and draw a region around the double negative (DN) fraction.
Step 2: Using the DN region from Step 1, gate a FS versus SS plot on these DN events, and draw a tight region around the subset of cells with low FS, now called Debris.
Step 3: Invert this Debris gate, creating a Not-Debris gate, and use this Not-Debris gate on the total population for your downstream Annexin V vs. PI display.
To see a screencast walkthrough of the gating process, watch the video below:
Let me break this down for you using an example. The entire point of the 3-step exercise is to properly define an area of the FS vs. SS plot that you'll call Debris, which will not be counted when generating your downstream frequencies. Depending on how much death you induce, and how toxic your treatment is, you could have a significant amount of Debris in your sample. It's important to note that you're not simply throwing away these events, you're just removing them from the downstream classification as Live (Annexin V neg, PI neg) Early Apoptotic (Annexin V pos, PI neg), Late Apoptotic (Annexin V pos, PI pos), or 'Necrotic' (Annexin V neg, PI pos). These debris events belong to none of these categories and therefore should not be included. The harm in including these debris events is that they usually end up in the Live fraction since they neither have a nucleus nor an intact membrane. Therefore, leaving them in inflates the Live fraction of cells, yielding potentially misleading results. The figure below shows the respective plots and regions.
Plot A shows a FS vs. SS plot of the full data set. Plot B shows the Annexin V and PI fluorescence on the entire data set with the double negative region drawn. In C, we are showing a plot of FS vs. SS gated on the DN cells. Here we can see we're left with two distinct populations. The larger population is the true Live cells (Annexin V neg, PI neg, and large), and the other population is the debris. I use a contour plot to show a nice outline of the region that should be drawn. This region is then applied to the top level FS vs. SS plot (as in D), and is inverted. This inverted Debris gate (i.e. Not-Debris) is then applied to the Annexin V vs. PI plot and quadrants are drawn to define the 4 populations present.
Not quite convinced yet? Well, I ran a similar sample on the ImageStreamX, gated in a similar way, but now, using the imaging power of the ISX, we're able to visually confirm what each of these populations is. To the left, we see the analogous plots and regions as prescribed in the 3-step method, and below we can see a representative pool of images from each of the regions. The panels labeled Debris and Live came from the DN gate drawn, and as in the flow cytometry data, the FS vs. SS plot (here presented as Bright Field Area and Side Scatter) separates into 2 distinct populations. A larger Live cell population and a smaller Debris population. Once the Debris is removed from the downstream classification, we are left with the Live, Early/Late Apoptotic, and Necrotic fractions. Of special note is the fact that we are not simply gating out small events, as there are small events that do in fact belong to the Necrotic or Apoptotic groups (see especially event #2269 in the Late Apoptotic panel). We're specifically gating out events that are small and have no fluorescence, which in this case, is a pretty good definition of debris. 
Now obviously there are a whole bunch of caveats. This works really well for a homogeneously sized cell line, and perhaps less well for a mixed population. Also, as was briefly mentioned above, you'll need to still factor in that % Debris metric somehow. It could be your treatment makes either live cells or debris, and excluding the Debris from the Annexin V plot may show you have a healthy profile. However, it's important to track % Debris in addition to the other populations when comparing your conditions. If nothing else, though, you now have a reproducible method for gating out debris from your apoptosis analysis that could actually be explained properly in the methods section of your manuscript.
