When a load (like your UM66) draws current from a resistor potential divider, it diverts current away from the divider, changing the currents in the resistors, which in turn changes the voltages across them. Consequently, a resistor potential divider cannot maintain a fixed voltage output under varying load currents, which will be the case for the UM66.
You can reduce the output voltage variation of the potential divider, under varying load conditions, by reducing the resistances in the divider, but you'll never eliminate that variation, and (worse) you'll have more current flowing, which wastes battery energy.
Here's a little demonstration of the problem:
simulate this circuit – Schematic created using CircuitLab
On the left (A) the divider produces an exact +4V output, as shown on voltmeter VM1, because there's no load current being diverted away from the resistors.
In the middle (B) I am pulling 1mA of current from the divider output, which causes a significant drop in output voltage, as shown on VM2.
In circuit C I mitigate the effect of load current by reducing the resistances by a factor of ten. This helps, VM3 shows a much smaller deviation of voltage from the desired +4V, but it's still less than +4V.
What we require is something that performs a similar function as the resistor potential divider, to obtain +4V (or whatever voltage you desire), but is able to automatically vary it's own effective resistance to compensate for changes in load current. That's the role of the zener diode that you mentioned having seen used in this application. Its resistance changes automatically to the right value in order to keep 3.3V across it regardless (well, almost regardless) of the amount of current taken away from the divider output.
I assume you are asking this question because you don't have a 3.3V zener diode. Fortunately there are other methods to achieve this. Since you have an NPN bipolar transistor, and I'll guess you have more than one, you can use them to make a device that behaves very like a zener diode. That's shown on the right here:
simulate this circuit
Instead of using voltmeters to show behaviour, I'll plot output potentials at OUT1, OUT2 and OUT3 as I sweep load current from 0 to 8mA. Below, the basic resistor divider output is blue, the zener diode circuit output is orange, and the transistor design is in tan:
As you can see, the resistor divider (blue) doesn't regulate at all. Its output falls linearly as load increases. The best behaviour is seen for the zener diode (orange). It's doing a great job of keeping the output fixed at 3.3V all the way up to a load of 6mA or so. Second best is the transistor design, which produces a respectable, but less "flat" output.
The best performance would be achieved using a linear regulator IC, such as the LM317L, the circuit for which would look like this:
simulate this circuit
This will produce almost exactly 3.3V for almost any load current, up to many hundreds of milliamps.