Accounting for confidence intervals in determining a mean

I have received some elemental analysis of a metal which I want to use to determine the effects of radiation. I want to report my results for the mean and 90% CL. The analysis only presents 8 data points for each element so I was planning on using the t distrobution as follows: M+(1.895)*(s/sqrt(N)). However, each of the data points is reported with its own uncertainty to account for instrument measurement statistics. These are reported at 1 sigma. My question is, in order to determine the 90% CL, should i apply the z distribution to each data point at the 90% CL and then calculate the mean from these?