I usually run at least enough exponential curve on the steering to make up for the positive mechanical expo you describe. Do what works for you!

You could use a setup station to measure the wheel lock versus transmitter wheel motion if you truly wanted to get as close as possible to a linear response. You'll never get it perfect, of course, because there's no guarantee the transmitter curve will match all of the non-linearities created by the linkages, and then of course there's Ackerman, so which wheel should you measure?