I think the glib answer is that they are two different machines. There are a lot of hardware differences between the two, ultimately affecting the sensitivity and ability to resolve the populations. Note, the older machine can only run in the 4-decade mode, whereas your Canto data is run in 5-decade mode. I'm curious if you calculate signal to noise ratio (stain MFI/isotope MFI) for CD25 to see which is more sensitive. The patterns look similar, but the Canto appears to have not resolved the CD25+ quite as well. This may be an optical illusion because of the different display modes. The percentages of CD25+ are also going to vary a lot based on the location of the gate, so that further adds to your differences.

Thank you Marcus, for your input.  I agree that the Canto population doesn't look as "resolved" as the Calibur population.  Could that be due to the bi-exponential vs log plot used to collect and analyze the data on Canto vs. Calibur , respectively?  I wouldn't think changing the gates would change the percentages, but it could affect where we draw the gates in the first place.  I'm interested in how to calculate the signal to noise ratio...does the software do that for you?  Also, do you have an opinion as to which instrument provides the most accurate analysis in this setup OR in general?  Again, thank you for your thoughts.

I have not used a Canto, but have lots of experience with the Caliber and an LSRII.  I would imagine that a newer machine would be made to be at least as good as an older machine, Then again, selections of filter etc that allow for more channels can affect the overall sensitivity and how parameters are displayed.  I've seen sensitivity and displays change with different filter set-ups.  

Signal to noise is simple for you to calculate:  Gate on the CD4+ and determine the Mean Fluorescence Intensity for CD25 expression.  Do the same for the isotype control (don't gate on CD4 if you don't have the CD4 vs isotype control).  Then simply divide the signal (CD25)MFI by the noise (isotype)MFI to get a ratio.  A higher ratio indicates greater sensitivity.  

Your data display differences, I think, is your major issue.  I'm hoping the S:N calculation will let you see which machine is more sensitive.  I have to say that I think the display of the Caliber data appears to be better for resolving the CD25+ subpopulation, but maybe this can be fixed on the other machine by changing the bi-exponential/log collection parameters.

My point about gating affecting the frequencies is not about the gates used before the plots you showed but, rather, the actual quadrants used.  My point is that CD25 expression is a continuum.  Moving those quadrant gates just a little closer or further away from the controls can have a big affect.  If your using FlowJo, you could use some of their algorithms to calculate the percent positive by subtracting controls from your CD25 staining on histogram plots.  That way no lines are drawn by hand and you'll have the percent positive calculated by math.  Having said that, I rarely use this technique in the real world because I find it to be more sensitive than I think is real.  However, it is unbiased – or consistently biased.

Whouldn't it be interesting to ask the manufacturer? I am not an expert but remember some measurements with FACScalibur and later machines. I am sure Marcus is right, that the newer machine should be better, but with more filters, I could speculate that under certain conditions of an experiment, such filters eliminate some signals which otherwise might be considered positive... but this is pure speculation, I am not an expert with those machines, but was fascinated by 16 colour staining in the late 1990s at Stanford.