Is it true that ...

Variable plots give "worse answers" for numbers of small trees?

No. Not in the logical sense. There is nothing about Variable Plot sampling that is unique in terms of the variability of the estimated number of trees.

Smaller Plots give less reliable answers. "Fixed" or "Variable" has nothing to do with it. If you look at a Prism BAF 20 (approx. 5 metric) and consider a tree of DBH 8 inches (approx. 20cm) you could calculate that "the plot" around that diameter class has a radius of about 15.5 feet^*.

Any plot larger than this will give more stable answers (in practical terms, although you could always construct a situation where this would not be true). A smaller fixed plot would give worse answers. It has nothing to do with the fact that you are using a prism. It is just a question of plot size.

At some point, the size of the plot around a particular DBH would equal a particular fixed plot, and for smaller trees the Variable Plot method would be more variable for smaller DBH classes (and better for the larger ones, of course). The break-even point would depend on the fixed plot size versus the BAF you were considering.

Which is best for the overall answer? Nobody knows, since there is no theoretical way to calculate this for practical situations. The belief that "fixed plots are better for numbers of trees" is just one of those urban myths. Maybe true – maybe false. If you wanted to, you could check it out for yourself.

You can easily calculate how variable your answers are by calculating the number of trees/acre at each sample point on some of your old data. You could then compare this with the variability of a fixed plot, (if you can come up with that data). Then you can at least compare one plot size to one prism cruise. Until then, there is just no way to tell.

^* The Plot Radius Factor for 20 BAF is 1.944 ft per inch of DBH.