People have said the same about the ancient Rover too... It just means you need to do more signal processing...
Electrical noise can be dealt with - I assume you're working from your own scratch built loom (A series minis not being known for EFI) and you don't have points etc for the ignition. And you'll have used shielded cable as appropriate. So as far as signals go, things should be as clean as they can be.
All that leaves is mechanical noise from the engine. Most of these should surely be proportional to RPM, so you should have some knowledge of what this will be at any given moment.
Record what a knock sensor "hears" on a PC and run it through a spectrum analyser and compare it with the RPM of the engine at those instants. Do a few tests at different revs and plot a graph of the main noise frequency peaks vs RPM.
Now compare these to your 1st harmonic knock frequency to see if they sit "in the way" at any RPM. If not, then you'll be able to do some good VEMS has a nice digital bandpass filter built in to the knock detection which can remove the noise.
If the knock frequency and the engine noise happen at the same frequency, then there's little you can do at this particular RPM with VEMS (remember the knock frequency is approximately constant, but your engine noise changes in frequency with RPM). With luck, this won't happen at high revs + high load where knock is most damaging.
There's a few games to play, but essentially once a signal is buried within noise that's similar in spectral characteristics it's the devils own job to get it out again.
