Luigi and the New Quest is a really nice SMB hack done by Darknight13.
btw, is Darknight13 still around? The "Nifty" board and website seems really out of date, and the e-mail there is invalid (an @home address).

I am working on an adventure game / scavenger hunt for some fo my friends.
The idea is that they would find what they think is a SMB1 cartridge, and
be required to play it in order to get the next clue. (For example,
embedded in the winning text on 8-4 or something)

I have created a modified SMB1 PCB that allows me to put mapper 0 roms
onto in using 2 27C256 256Kb EPROMS. (I put the PRG rom on it directly,
and I make 4 copies of the CHR and put it on it)
I have now had success with "SMB1", "burgertime", and "duck."

However, when I tried to do this with LNQ, it seemed to be
fine, the graphics looked fine and it started ok, but the default level
was not 1-1 but rather 0-1, and when you play it it looks like you are on
some random level in the code; it is underwater, no enemies, and after a
short while you get to a bowser bridge and "win" the game.

I have tried patching a number of different versions of SMB, it all seems the same.

The same source rom plays fine on NNNesterJ, VirtuaNES, FakeNES, and FCE Ultra.

I did some more extensive reverse-engineering last night, and it seems as though the starting level is encoded in a different way on LNQ than on SMB, so game genie/hex edits that change the start world on SMB do not work on LNQ. It seems as though the level data is in the same place on the ROM for all levels EXCEPT 1-1, though the 1-1 enemy information is in the same place.

This is interesting, since it seems to be a rather endemic bug in the emulators in that the game plays different on actual NES hardware.

Any ideas? Is there a way I can hard-code the starting level?

I'm having the same problem that you described on the FCE Ultra emulator.
I patched a rom that NEStoy renamed to Super_Mario_Bros_(No_Title_Hack).nes. When I went to play it in the emu it starts on world 0-1 as you described.

I was hoping to play the game with IBMNes...

I had this problem a while ago after recompiling a hacked version of smb1. When I ran it under my emu and LissNES it seemed to work fine, but once I loaded it in NEStopia and FCEUltra the game would start off at this level. It turns out that these hacks do not initialize the RAM, and (apparently) the game later on reads memory spaces from RAM and expects them to be set to specific values. Thus, this hidden world is shown... I am guessing that the only way of fixing this glitch would be to modify the RAM location used to store the current world number accordingly.

hydesprojects.cjb.net

I wonder if that would be possible with a game genie?
I know that the first poster mentioned that he had already tried a game genie to no avail. Maybe I'll just give it a shot for laughs. I might also try to find a different version of the original rom to patch.

I'll let you guys know if I make any progress.

As a side note... has anyone emailed the author of the patch about this problem?

Well, to be honest with you I do not know how game genie works, but I am sure that there are other ways of modifying specific RAM locations. What you also might wanna try would be to disassemble the game, modify it so that it looks like a mapper #2 game (the one with PRG pages only) and add RAM initializing code to, say, a third PRG bank. Accomplishing something like this would be hard, but doable. It would also allow you to add even more functionality to the game (perhaps more graphics?). I imagine this would turn out to be expensive, but oh well...

I don't think anyone has notified the hackers about these problems yet. With so many smb1 hacks out there, it's hard to find the people who worked on them. Besides, I am sure the authors themselves are aware of most of the problems found in their hacks...

hydesprojects.cjb.net

I did try to get in touch with the author of this excellent hack, but it seems like his website was pretty abandoned last I checked.

An interesting thing to note is that this works on certain NES systems but not others, presumably based on HW revision or something. And it is consistent. So if it is caused by not initiliizing certain RAM locations, the power-on or reset RAM contents in question are consistent for a particular HW revision...

For my project, I just tried about 5 NES systems and it worked on 2 of them.