Right, so the reasonable inference would be "this proxy can't event match the temperatures we know for sure -- it's no good, throw it out entirely".
The proxy *may* be fundamentally flawed and the modern divergence from instrumental records *may* be a clue to that flaw, but you do not know that.
What you do know (if you were Keith Briffa, rather than some guy on the internet second guessing him over a decade later) is that the site was carefully selected so that the growth of trees there would be temperature constrained, that the proxy matches well to instrumental temperature records prior to the divergence point, that outside the instrumental era it matches well to other proxy reconstructions going back for hundreds or thousands of years and that these other proxies *do not* diverge with modern instrumental temperatures in the way that this proxy does.
So a reasonable conclusion is that there is a confounding factor which has superceded temperature as the limiting factor on tree growth from 1960 onwards.
Now it's possible that the confounding factor is one that is 'naturally occurring' and thus could have polluted the record at random intervals going back throughout the entire duration of the data series rendering it useless as a proxy, but if that were the case you would expect to see declines in this proxy which don't match other proxies (unless those other proxies are also affected, but then that begs the quesion why aren't these other proxies affected today?).
A more parsimonious explanation is that the confounding factor is modern in origin and that the data series is perfectly fine as a proxy up until the modern era. Given that we know that modern industrial civilisation has had massive impacts upon the biosphere, the idea that there is something about the modern world that has hit these trees in the last few decades and rendered them useless as proxy thermometers isn't especially outrageous.
But what do you do whilst you try to figure out what the confounding factor might be? Do you sit on the data even though you have good reason to believe that it's OK prior to 1960? Or do you publish it with appropriate caveats and warnings?
Keith Briffa took the second approach and I, for one, think that it was both eminently reasonable and the scientifically proper thing for him to have done.
Regards
Luke