How to combine instrument uncertainty with deviation

[Celder]

Here's my problem: I have been using a digimatic indicator tool to measure the thickness of a somewhat uneven film. The digimatic indicator has a digital screen that shows up to 1 um, but according to specifications the accuracy is only 3 um. To account for the uneveness I have taken 10 measurements at different points of the film.

My question is: When I report the average thickness, how should I calculate the error?

 

[Simon Bridge]

Welcome to PF;
Check the machine spec to see what they mean by "accuracy".
Rule of thumb is to take the uncertainty on each individual reading to be half the instrument resolution.
It is not unusual for you to have to use equipment which has a higher resolution than the uncertainty in the measurement - i.e. you cannot use a stopwatch to 1/100s accuracy but it displays to that resolution.
What do you normally do with the stopwatch measurements?