While the current models certainly represent a vast improvement over the older codes, unfortunately, the new models may not give an adequate representation of reality. Arimoto et al. (1997) suggest that inaccurate atomic physics may be responsible for the derivation of unphysical abundances in elliptical galaxies, and discuss possible problems with abundances derived for other sources as well. They find discrepancies in abundances derived from different models of factors of 3. While such large discrepancies may reflect the uncertainties in the quantities calculated there is no way to determine the systematic errors in quantities currently ignored by all models. Only high spectral resolution can tell us what we have missed.
For some ASCA observations no acceptable fits have been found. In particular, Capella has eluded chi^2 <~ 6 since the PV phase data were collected. Given the observed emission lines in the simultaneous EUVE spectrum, rather tight constraints have been imposed on what the problem(s) could be (Brickhouse et al 1997). Liedahl & Brickhouse (1997) have finally come up with a possible atomic physics explanation (which remains to be tested), the implications of which are rather far-reaching.
In short, models derived from ASCA observations of supernova remnants, cool star coronae, cooling flows, clusters, starburst galaxies, and galaxy groups are all under ``suspicion.'' Observations of the hot interstellar medium with other instruments (e.g. DXS, Sanders et al 1993) have also raised questions about the adequacy of the atomic models.
From the plasma spectral modeling point of view, all collisionally ionized plasmas need similar types of accurate collisional rate data. Thus, benchmarks derived from stellar coronae will have widespread applicability to other sources. Of most concern are extended sources, for which grating data may be difficult or impossible to deconvolve. While the stellar coronal sources themselves may be diagnosed using presumably well understood, strong emission lines, spectral observations of extended sources will continue to rely on global spectral fitting techniques, and will place a far greater demand for accuracy and completeness on the spectral models.