OK, but suppose you ignore provinces all together and force the Smithy BMU player to buy 8. How long does that take? How much are silk roads+ islands + some duchies + estates worth by then? I don't think it's obvious. At least the province player needs to be rushing the game with a salvager or something.
I tried it. It's a job for the simulator, but I did it by hand a few times... BMU-smithy, buying only provinces, ended the game somewhere between turns 20 and 28, with 8 provinces and 3 estates for 51 points. So going by the points carlos had accumulated, (53 by turn 22), it's close. Whether BMU+s could end the game on turn 20 or 28 depends on shuffle luck. The point total for BMU+s would be between 51 (no duchies) and, realistically, probably 60-63 or so (doubt you'd fit more than 4 duchies in). Whether the BMU+s player got duchies at the right time or started going for them too early or too late would determine whether the duchies would slow down him ending the game on provinces or not.
Of course, given that this hypothetical BMU+S player is not competing with you for either Silk Roads or Islands, the non-smithy player would be able to keep his deck a little slimmer (more islands) and higher-value (more silk roads, which were worth almost as much as provinces by the end of the game). Those silk roads were worth 5 - only 1 less than a province! Silk Road + island was worth more than a province.
It's close enough, I think, that I'm not comfortable saying "BMU+smithy is better than silk_road-island-pawn" or vice versa, without going into more detail.