The correlation looks perfect. Is the candidate method acceptable?

A new chemistry analyzer is going in beside the one the laboratory has used for years. Both run the same hexokinase enzymatic procedure for plasma glucose, and the validation team splits paired fasting specimens across the two instruments. The printout comes back with a correlation coefficient of r = 0.999. Someone on the bench says that number settles it, the candidate method is fine. It does not settle it.

Correlation describes how tightly paired results cluster around a fitted line. It says nothing on its own about whether the two methods would report the same number for the same patient specimen. A candidate method can track the comparator almost perfectly and still read a few milligrams per deciliter higher at every concentration, which matters at a diagnostic cutoff even though it barely dents the correlation coefficient.

Which calculation or graph actually answers whether these two methods agree closely enough to use interchangeably, and what can a regression line not prove no matter how good it looks?