grouse1,
EdStainless is correct, there are ESR grades that will have fewer inclusions, but I am not sure that what you are seeing is substantially affecting the fatigue performance. Unless they are excessively clustered, it is normal to see small, elongated nitrides (TiCN, AlN) in 17-7 spring grades. There was an excellent article by Mark Hayes of IST in the November 2006 issue of Wire Journal International (originally presented during the WAI conference in Prague, 2005) that addresses this very point. The following is an excerpt:
Any metallurgist who has looked down a microscope at a longitudinal section taken from a 17/7PH spring would not describe the steel as clean. The number of inclusions is considerable, as shown in Fig. 2. However, close observation shows that these titanium carbon-nitride [sic] and aluminum inclusions are mostly in the size range of less than 10 microns. IST believes this is a full explanation of why this grade has such good fatigue resistance when springs have been glass-bead peened.
This last point is key-- surface initiated fracture is typical in helical compression springs, hence shot peening is a typical process for highly-stressed applications. For sheet-based springs, peening is unusual, and therefore the surface condition (roughness, defects, residual stresses) is very important for determining fatigue strength. I would evaluate these aspects before concluding that inclusions are the primary concern.