I did not cheat, but made the requested consideration before reading on. I concluded that I accepted his tests called Condorcet, No-hopers and Steadiness but I totally rejected his test called Non-transferables. I then found, not much to my surprise, that he had found Meek's method to be a clear winner (on his particular data) on the three tests that I accepted as valid, while Newland and Britton (2nd edition) rules had done 'better' on the Non-transferables test which I had rejected, so I think it important to explain just why I had rejected it.
My view is that everything should always be in accordance with what the votes say, in proportion to their numbers and, if some votes, in whole or in part, are entitled to transfer and do not indicate a wish to be transferred anywhere, then it is morally wrong not to make them non-transferable, in whole or in part as the case may be.
That being so we cannot say which of two methods is better on the basis of the number of non-transferables, until we know the cause of the difference. If method 1 shows more than method 2, we must ask whether this is due to method 1 making some votes non-transferable unnecessarily, or to method 2 failing to make votes non-transferable when they should be. With methods of which we know nothing except the outcome of this particular test, we can really say no more than that.
In the actual case, however, we do know the methods in detail and are aware that Meek's method never makes anything non-transferable except when it is right to do so. It follows that, if the Newland and Britton rules get a smaller number, it is they that are failing to do the right thing.