1/2) More or less, I try to clarify, It seems to me you believe your graph in figure (2) shows R and L variations w.r.t frequency.
That graph simply shows impedance of a parallel resonant circuit made of approx constant 100nH inductor, 45kohm resistor and a \$\frac{1}{(2\pi\times 2\,\text{GHz})^2\times 100\,\text{nH}}\approx 65\,\text{fF}\$ capacitor.
This is one first level approximation which may however be good in many circumstances.
Then one can argue resistance is frequency dependant and add this to the model, the same for inductance and capacitance but this can usually hardly be seen on graphs unless you carefully measure, analize and fit curves to models.They are ususally hidden in measurements errors.
Then you can add many other extra parasitics and non linerities outlined above by other contributors, but there's no such an evidence in the sweep you posted.
b) Your meter shows variable (possibly negative) inductance just becouse you ask it to do so. It just measures an impedance, one complex ratio between voltage and current and then represent it as you configured it to do: you asked for Ls+Rs and if measured impedance phase does not agree with that model it just keep computing and findes "negative inductances".
Usabilty boundary (green/blue transition in fig. (2)) is not function of your component parasitic only, but it depends pretty much on the rest of the circuit e.g. if using that inductor in a resonator you should add stray capacitance to the computations and see if you get consisten numbers.
4) Yes SRF is that blue to red boundary.
3) Dimension do count. For resistors and capacitors usually the longer the part the higher the stray inductance. E.g. small SMD 0204 or 0603 may exhibit a few hundred pH while some big HV MKM capacitors I happened to use were specified as 7nH/cm w.r.t to terminals pitch.
I have also experienced the same humming noise while using PWM dimming on Fan. The only solution that works well is by using capacitors to reduce speed. You may need to stack 2 or 3 capacitors (1uF, 2.2uF, 3uF) which are rated to work in 250VAC and connect them in series to get various speed levels. It is a little complex and bulky compared to PWM approach, however this will not create any humming noise.
Best Answer
The table says this:
Self resonance frequency (min.) 26MHz
Notice that this is a minimum figure. The graph shows this: -
And the graph is a typical graph (although it doesn't state so). It's pretty normal for graphs in product data to represent typical values.
Because the table indicates the minimum SRF.
If you dig out the data sheet you'll see that the 10 uH and 15 uH (yours) have the same stated SRF: -
Hence Murata are likely being a tad lazy with their specifications.
You can't unless you want them to behave like capacitors.
I think Mini-circuits might be mistaken here but, contact them to find out how they can justify this. Alternatively, the losses may increase so much that the device becomes very resistive at higher frequencies maybe like a ferrite bead: -
Image from here.
Although the SRF is about 60 MHz, the impedance continues rising (due to losses in the ferrite material) up to a usable point beyond 1 GHz. It's not the way I like to design things but you can't always rely on someone else's design matching what you might do. In other words, if an inductor isn't characterized at some arbitrarily high frequency then don't use it. Don't even use it if you have tested a dozen and find them all to be good at the high frequency because how would you ever find a replacement if you needed one?
It's like finding some old paint in your garage and painting your kitchen. After the third wall you run out of paint and, of course, the colour matching process to finish the last wall will be a pig of a job. You'll probably end-up starting all over again and curse yourself and your dog and anyone else who happens to be in the vicinity but, who really was to blame?