I don't know that the 15% invalidation rate reported by the Journal Sentinel is the right one. I do think that it is very unlikely that recall challengers will invalidate enough signatures to avert a recall. I don't recall talking to many politicos who harbored much doubt that public employee unions could gather enough signatures to force a recall.
The paper's method of selecting petitions, developed in consultation with MULS Visting Prof Charles Franklin, seems reasonable and my understanding is that the margin of error is plus or minus 4.5% That margin might be larger if the rate of invalid (or unverifiable) signatures is not evenly distributed among circulators - as it might not be if invalidity is substantially impacted by intentional misconduct. If there were a small number of circulators who turned in a disproportionately large number of invalid signatures, that might affect one's confidence in the estimated strike rate. But, again, my understanding is that the more petitions are spread among a large number of circulators, the less likely this is to be a problem since all circulators had a chance of being included in the random sample in proportion to the amount of signatures that they turned in.
In addition, the actual rate of invalidity may differ depending on the method of verification. If, for example, the Verify the Petition people use a more exacting verification process, they might produce a higher rate of invalid signatures.
So I don't think there are a million valid signatures but I am pretty sure that there are more than 540000. There is, nevertheless, value in vetting the petitions. The public has an interest in knowing whether there are really a million signatures and, if some circulators or signatories broke the law, that fact is significant even if the number of signatures exceeds the minimum.