Thank you for your reply and pointing out this issue.

An alternative is to use PRISM and so calculate the amplification efficiency ± CI. As Mikael points out, this can be quite an eye-opener. For one of our RNA-based dilution curves what was an acceptable slope of -3.389 turns out to actually range from -3.990 to -2.788, which of course needs to be accounted for when calculating copy numbers. And if you do a relative quantification, then you must make sure you carry over your errors appropriately.

We've taken to running standard curves multiple times using pooled samples from the set to which we'll be applying the efficiency correction. Theoretically this should account for both the primer efficiency and any sample interference (though it still misses any sample-to-sample variation). We then average the slope over multiple runs and use this number as our efficiency correction. While the confidence intervals from one dilution curve are fairly broad, we actually see fairly minor experimental variation in slope over the repetitions.