I do not understand why the smaller nanoparticles have the largest IP (even 0.7) and have not found an equation that satisfies the values of IP that gives the Zetasizer equipment.
The zeta sizer use a basic cumulant analysis procedure. The means the log of the autocorrelation function is fitted by the polynominal function a+bt +ct2. Pi=c/b2.. For monodisperse particles this value is zero. CONTIN is to prefer if possible.
As mentioned, the Zetasizer uses a cumulant analysis method.
In answer to why you seem to measure a higher polydispersity for small particles, Dynamic Light Scattering is less sensitive to small particles as the scatter less. This means that your measurement will be sensitive to dust or aggregates which will lead to a high polydispersity value.
for broad polydispersities see the recent article by Mailer et al and references therein.
A. G. Mailer, P. S. Clegg and P. N. Pusey, J. Phys.: Condens. Matter 27 145102 (2015). Particle sizing by dynamic light scattering: non-linear cumulant analysis
The polydispersity index or PDI is sometimes also called the normalized second cumulant defined from the polynomial fitting parameters. For a Gaussia size distribution this would correspond to the square of (the width divided by the mean). Large PDI values indicate that a cumulant fit is not properly describing the situation in your sample, and a regularization may be a better representation of what the situation. Yes, smaller particle distributions are more sensitive to dust and aggregates (centrifugation and or filtration may help) - and also if you are preparing these yourself, larger particles grown in stages tend to turn out more narrow.