I listened to your sample clip and believe I understand what you are trying to say - that the noise is most likely inherent to the environment. However, in its current state it is too much in the foreground and is actually obscuring the dialogue. The bad news is that there is one specific whining noise that is right in the frequency range of the dialogue and any noise reduction you do is going to bring that forward along with the dialogue. The good news is that you should be able to push most of the broadband noise back enough to make the dialogue intelligible.
I ran your sample through iZotope RX and was able to run 2 passes reducing the noise by 8dB each pass without creating noticeable artifacting. You should be able to achieve much better results using Waves Z-Noise, as you will be working directly from the WAV file.
Start by capturing a noise profile from a portion of the recording where there is no dialogue using the "Learn" function. Then, once you have pushed the environmental noise back sufficiently, capture another noise profile, this time of the dialogue, using the "Extract" function. You should, with a bit of experimentation, be able to reduce that whine sufficiently that it is not intrusive.
A few important things to remember:
1) You will achieve more natural results using 2 passes reducing the noise by around 8dB each pass, rather than 1 pass reducing the noise by 16dB.
2) When running multiple passes, either use the "Adaptive" function in Z-Noise or obtain a new noise profile before each pass.
2) Listen to the difference before running each pass to ensure that you are not removing dialogue and, if you are, try reducing the threshold slightly until it goes away.
2) There is always the temptation to reduce the noise by too much. Listen carefully to your results after each process and when it starts to sound artificial you've gone too far and need to undo your last process.
Hope this helps!