Yes. I think you may be thinking that the errors in estimating standard errors could be high, and I agree, but the estimated confidence intervals are going to at least warn people by their approximate length. Otherwise, many may look at a point estimate and may not think much about the error at all.

I assume you need higher accuracy for your study.

You might, if you think it instructive, have an appendix showing a sensitivity analysis of sorts, showing how leaving out one or two observations could change results, and thus speculate about the accuracy of the confidence intervals estimated.

Perhaps a bigger problem is violation of your assumptions. If you are assuming random sampling, that is often violated, even when planned well, by nonresponse. Imputation can be tricky. And if you are using models for prediction (i.e., estimation for a random variable), then you still should be aware of any potential problems, including having appropriate and accurate predictor data.

Depends on your study objectives, really. Both statistics are used for different purposes. Percentage is a point estimate for the sample of the data. Whereas, confidence interval is an estimation of this point in the sample inferring to the study population. In short:

percentage = for sample

confidence interval = for study population

The distribution of the sample of each group can be assumed to be normally distributed when you have a set 30 sample minimum following the Central limit theorem. But still,if less than 30, you can check using histogram to confirm normality of data before proceeding with the analysis.

Hope this helps.