I guess it would be enough to just take the Maxim sensor and present the temperature with one decimal and don't think twice about it. The customer don't know what he needs/wants except that it should look good. Why else require that it matches the reading of a Fluke?

And if you use a DS2438 Smart Battery Monitor, you get a resolution of 1/32°C. So cool to show the temperature as 11.3125°C. After all, if you do got the digits, what reason is there to wonder about that 0.5°C measurement error within the narrower temperature range or 2°C at the end of the measurement range?

The advantage with a Dallas sensor is that it has quite small drift from aging and already produces a digital result. If you want to, you can apply a correction table to "draw" the measured value closer to a reference thermometer if the goal isn't to be correct, but to have multiple thermometers show very similar results. In some situations, the repeatability is more important that the absolute tolerances.