I need more help with figuring out how to weight my data points.

It isn't easy to choose a weighting method, and most people never deal with it.

It is tempting to weight points by the scatter of the replicates, giving a point less weight when the triplicates are far apart. But unless you have lots of replicates, this doesn't help much. The triplicates constituting one mean could be far apart by chance, yet that mean may be as accurate as the others. Weighting needs to be based on systematic changes in scatter. If there is more scatter with the points with larger Y values, then you want to use weighting, but it is hard to know whether this is happening.

What happens if you choose the wrong weighting factor? Not much. At least for the cases we have studied by simulation (dose-response curves, exponential decay), the results are only a bit less accurate if the weighting scheme used to analyze the simulated data doesn't match the scheme used to generate the data. But the difference is pretty subtle.