Earlier quoted context omitted.
OTOH the first log table was computed by taking successive square roots of 10. Then you use the binary expansion of each number on your scale, right? So compass-and-straightedge doesn't seem crazy, since square roots are easier that way than numerically by hand.
The original question seemed to be asking about how to make a real chart, that is, with real paper, pens, compass, straightedge, etc. In the real world, taking successive roots is going to bleed accuracy pretty quickly, and when you feed that inaccuracy back in to the e^x function, it's going to bite pretty hard. I don't think it's practical to get the requisite level of accuracy this way.
Yeah, I don't know how practical a construction would be overall for someone who seriously went at it; I was mainly moved to answer because this objection that constructions can't express transcendental numbers just doesn't seem relevant -- digitally you don't keep infinite precision either.
Square roots and multiplying are two of the simplest geometric constructions (geometric mean and taking a proportion), compass and straightedge operations can be accurate, you can construct it enlarged and then scale it down, and finally the precision you need to aim for at the end is bounded. A potentially fun project idea. Think of it as a kind of retrocomputing: how might Euclid or Archimedes have designed a slide rule?