There are two different problems related with the temperature structure of H II regions that are still controversial: a) typical observed t2 values are in the 0.01 to 0.04 range, while typical values predicted by CHPM's are in the 0.005 to 0.020 range. The differences in t2 between CHPM's and observations of H II regions are smaller than in PNe but I think that they are real (see the review of Esteban 2001), and b) in general photoionization models predict T(O III) values smaller than observed (Stasinska & Schaerer 1999; Luridiana, Peimbert, & Leitherer 1999; Luridiana & Peimbert 2001; Luridiana, Peimbert, & Peimbert 2001; Relaño, Peimbert, & Beckman 2001) indicating the possible presence of an additional heating source not considered by the models.
The warning mentioned by
Viegas (2001)
regarding point b) above should be
considered: models depend on many assumptions and a very good model is
needed
before we accept its implications. For example the difference between the
observed and predicted T(O III) values depends on
the adopted filling
factor, 
the difference
between the observed and predicted
T(O III) values for I Zw 18 disappears for models
with values of > 0.3.