Hi mir4ea, I would myself probably go with the filleting method, just because it can generally be difficult to control the surface shape in pinched-down areas.

But it also depends on what in particular you want to control - the filleting method lets you control the end cap shape more easily since you can build a more simplified surface for it but you are relying on filleting to generate some pieces for you instead.

One thing that can sometimes help with using Network for "collapsing down tip" areas is to put in a small circle or arc curve near the tip, that kind of tends to make the shape even out somewhat as it comes together, a couple of examples of that here:
http://moi3d.com/forum/index.php?webtag=MOI&msg=3623.14
http://moi3d.com/forum/index.php?webtag=MOI&msg=3550.6


In the end, there is not any "one absolute right way" to do stuff, it can sometimes just depend on your personal preference.

Just in general though I find it difficult to get good surface quality when you're trying to force the surface through too many different shapes all in one single surfacing attempt. So building some broad forms separately and using fillets for transitions can tend be good instead of trying to model them all at once.

Also if you need to make a model that is sort of blobby / melted looking stuff like that can tend to work better in a sub-d modeling program rather than in NURBS.

- Michael