I have recently found the same problem with photosynthetic rate data
measured on single leaves of tall fescue. In my case, the use of
non-parametric analyses (one option) was also not possible because I had
several treatments, in effect a three-way ANOVA was needed. I am reduced
to reporting untransformed means and ranges and being very careful about
the nature of any statements of supposed significance that are made.
I too would be interested in any suggestions that other readers of this
newsgroup have to offer.