The green wire is set to every LM3916’s pin 5. Sorry, i know its kinda hard to see. But pin 5 on the LM3916 is signal, so they should all be tied togeather.
The Blue wire is going from LM3916 pin 1 to the led bar graph pin 1. This allows 10 leds when the chip only has 9 pins.
Cascading a LM3916 is likely to not produce the results you are looking for unless that is non-uniform steps. For cascading, the LM3915 is a better choice because it has uniform 3dB steps.
But there is a reason why the device data sheets do not cascade devices the way that the SparkFun page suggests. To illustrate this consider the voltage divider of a single LM3915.
If Vhi connected to some reference voltage, Vref, then the voltage at the input of comparator at pin 11 is Vref * 16K/(16K+6.63K) = 0.707Vref or 3db down.
Now cascade two and compute the same voltage divider output for the top LM3915:
Vref * 38.63K/(38.63K+6.63K)
Which is very different (-1.3dB) and probably not what you wanted. With four units cascaded, it gets even worse. I suggest that you extend the opamp based method of cascading as recommended in the device data sheet. But four units is probably not worth the effort because it will be impossible to get anything useful out of the lowest unit. It is working at -90 to -120dB which is likely to be in the noise not to mention the offset issues.