For heaven's sake. Talking about the good old semiconductor transistors, Does this mean that transistor books never mention about a transistor being linear at all? Of course it does! If you will take into account the dark saturation current and all its external parasitic capacitance, inductance, internal parasisitc shunt and series resistance, temperature, minority carriers due to imperfections on manufacturing .... then you will indeed have a non-linear curve at a very close investigation of the curve. If you look into any semiconductor device at a atomic level then it will be indeed non-linear. But these very very small non-linear response can be ignored in most cases and that's what most electronics book authors agree. If a very simple question is raised by an inquirer and answered by all of these factors that affect linearity then the inquirer will have to understand the very mathematical physics of semiconductors. You said it yourself that "It's just sometimes more linear than other times", and that is at the ACTIVE region and most transistors are operated at this region. It's a simple question that only requires a simple answer.