I'm also a strong adherent to the top down approach and as Tuura I really prefer myself and I advice my players to start with the flaws. But this is also a matter of playing style - in what kind of stories the character will be involved. I prefer stories that build heavily on the characters themselves which in turn requires the "Solid Character Concept" to succeed.
A possible challenge is that flaws might be a very big chunk to swallow, but I do feel that most make their point without neccesarily knowing the rules too well. But having played the game - like all rpg mechanics - will always give you an advantage. Both in terms of mechanic minimaxing and in terms of story/roleplaying potential. Basically I just think that the flaws are what defines a character the most and that if you havent yet any concept in place the choosing of flaws (and virtues) forces you to start making decisions as well as inspiring them.
These choices are the most important for the magi characters - bc the rules have a big impact on their hopefully long life and their success in the lab. And because they will be a center of affection for most players. Since magi tend to live long and be among the first characters made this poses a problem when people new to Ars starts up. My choice in handling this has been by starting my current saga - where ars really was all new to most of us (some had tried a little peek of 3rd ed Ars several years before and some had played a little bit with the 4th ed Ars but in a viking setting with no magi... yet) - by letting all the players make apprentices. They played apprentices for a year and half (14 years ingame) and some of the final flaw-choices where postponed till later in their apprenticeship. New players have joined us and some of them have started their magi character out as apprentices too. Besides the character generation issues this has also give other benefits: it has been a easy easing into the setting as well and we have had a lot of fun with stories, scenes and concepts only really doable with apprentices.
Finally I dont completely write off the spell/bottom up approach - but rather actually as a part of the "Solid Character Concept". I am dabbling a bit with making a magus myself - in hopes of getting more active beta-SG - and this character concept is completely dependent on being able to do certain effects. In a sense this makes the char gen come full circle. And I think that no matter what approach you use there is an important point in your last paragraph: that one's ideas will have shifted by the time a char gen is at its end. And as a testiment to that I can recall few char gen I have made in any system where I hadnt at some time gone back and changed or adjusted previous choices when getting to later choices in the char gen, no matter the order of approach.