Even if this is just a illusory construct, the thing generating the illusion must still exist. (i.e., if all this is just a dream, the Dreamer is still real.)
At some point somewhere, something has real, undeniable Existance. What the nature of that Existance is, and how that’s relevant to me, is another story altogether. 
Even in a purely probablistic context, the Turing test can be wrong too often. It’s simply not reliable unless you have a very astute test-giver, and even then, it’s only useful if the test-giver is very familiar with methods that the object might use to try to trick him. What good is a test that depends more on the test-giver than the object being tested? Repeatibility should be a part of any scientific experiment, and the Turing test can easily fail as soon as another scientist does the testing. I suppose the way out of that would be a mass Turing test, in which X number of people have to vote for sentience in an object before it’s “official”. But sentience by committee? I don’t know, that just seems like more of a lazy, pragmatic solution than a real test. Also, it still doesn’t deal with the fact that the Turing test deals with human gullibility and the ability of objects to deceive us – which is not necessarily the same as sentience, even in principle.
But really, what’s the point of the Turing test anyway? Can’t something be humanlike but not sentient, or perhaps sentient and not humanlike? The Turing test is an attempt to confound the two simply because we currently lack the ability to truly measure or define sentience.
Once we figure out what this mysterious “sentience” is, proper tests will naturally come.