I agree with Vladimir that negative stainig will give you the information about bacteria shape, presence of flagella, fimbriae or capsule. However, if you are not working with pathogens, I would suggest to omit the fixation step at all. The fixation breaks down the permeability barrier of bacterial cytoplasmic membrane. The negative stain then easily penetrate into the dead bacterial cells.
In our hands, 1% ammonium molybdate pH6.8-7.0 with 0.1% of trehalose works well for negative staining of unfixed G- bacteria.
For a fine TEM analysis of bacteria (capsule, flagella, plasma membrane and inner ultrastructure), a good fixation, graded dehydration (alcohol), propylene oxide, resin embedding, ultramicrotome sectioning and heavy metals salts stain, are mandatory.
I completely agree with you. I think that negative staining is the method of the first choice. It is fast and it can save a lot time in preliminary characterization of unknown bacterial samples, e.g. soil isolates. However, it cannot replace the classical ultrathin section analyses in any way as you point out.