Every rule applies poorly to some corner case. That is a necessary consequence of being simpler than reality: Every time you simplify a model by disregarding a rare or unimportant trait, you make the model poor for the cases where that rare/unimportant trait applies. And you have to simplify, to avoid an unmanageably complex rule. (Let me tell you about the German tax code some day. They really tried to be fair and didn't value simplicity or comprensibility highly.)

In the case of this tree, the computed value according the rule is a very high value, and the real-world value is also a very high value, so what's the problem? This is a rule that produces a fairly good result even in a case where it may be expected to apply badly. That's not a bad rule.