Your analogy isn't really clear and not the best to compare with my example. A deformed ruler can yield consistent results, so by your definition, it is reliable(?). Sure, reliability is about consistency, however, it must also be trustworthy results (well, they are synonyms of each other). A ruler which is deformed is not trustworthy for this experiment within the reasonable error bounds, hence the results, whilst consistent, are not reliable. Can you really trust results which use a deformed ruler? No. You raised the point about accuracy, but all experiments are not 100% accurate; inaccuracy and unreliability are not mutually exclusive.
Anyway, there's no point debating this if it's coming down to how we define the word - Look, I could be completely wrong and completely off the mark, but I've written what I said numerous times in many reports both scientific and historical. It hasn't been raised, so there's no reason to alter my definition.