THx for the replies,
the reason i asked here was that given such argument, why students (and those published in many IEEE papers) are still struggling with getting a 'good return loss' for the antenna alone? why not design an antenna by simply looking at the radiation patterns and match the antenna using MN afterwards?
i could only think of the following:
1) insertion loss in the matching network,
2) bandwidth requirement,
3) cost of the matching networks (caps, inductors for <6ghz, or stubs for >6ghz)
is there any other concerns about getting a good return loss for an antenna?