I disagree. Unless we are talking about sofware engineering then it seems to me that what you select is based on previous projects but the choices themselves are based on tested scientific models with predictive power.
To clarify; the use of precedent in engineering is not objectionable (on the contrary, it is quite sensible); it merely runs counter to this popular idea that engineers are forever deciding everything via Science.
You seem to be saying that any engineering precedent must ultimately be based on a scientific model somebody used in the past. Well, maybe... if you're willing to call "we tried it this way and it seemed to work" a scientific model, then okay.
Rationality quotes time!
The usual rules: