Edit: This was posted before reading Barium's reply.
The reason you have a problem with MDMA formation is because the N-OH group is far less stable than a typical C-OH group (i.e. an alcohol). N-OH is readily reduced to N-H by a large number of reducing agents, and possibly by other means (excessive heating, for example). The N-OH group is so readily reduced to the amine that plain hydroxylamine is often recommended for the synthesis of substituted amphetamines; most reducing agents will give nothing but the amphetamine when performing a reductive amination of a phenylacetone with hydroxylamine.
With the more specific reagents such as sodium cyanoborohydride you can halt much of the reaction at the N-OH stage - but, as I said above, there will always be the chance that you have reduced some to the amine. Further, you have no way of knowing, and you cannot take for granted the stability of the product even after the reaction is finished. It may decompose during workup, or during distillation. I want you to be sure of the potential risk you are taking, so I will quote from obia's post
Post 324207 (missing)
(obia: "UK analogues", Law and Order). I hope you can appreciate the relevance of the following:
the defence case was that the defendant was making n-oh 2cb (which wasn't and still isn't illegal despite the trial judges numerous attempt to rule that it was) and if the substance was in fact 2cb containing then this was either accidentally created by a botched synthesis or was an artifact of analysis.the prosecution case was based on dubious assumptions regarding the stability of n-oh 2cb particularily to heat. however the jury were so stupid that they couldn't follow the complexities of the chemistry and convicted anyway simply because the simplistic prosecution case was easier to understand than the defence case. it has since been shown that the product of the synthesis is indeed almost pure n-oh 2cb which decomposes during the final workup used into 2cb and an oxime. further more there remains no trace of the oxime in the accidentally made 2cb (this was vital to the prosecution assumption that the product was deliberately synthesised 2cb and not accidentally synthesised from n-oh 2cb).
To answer your next question, sodium triacetoxyborohydride (STAB) can be readily made from not-particularly-toxic sodium borohydride and acetic acid, see Post 439071
(Barium: "Prep of Sodium Triacetoxyborohydride", Novel Discourse). But the suggestion - even by Rhodium 
- that it is very likely to selectively reduce to the N-hydroxylamine does not mean that it cannot fail. Other reagents that may work are plain sodium borohydride under controlled conditions (e.g. very low temperatures; Shulgin N-methylates MDOH to FLEA with formaldehyde 'at dry ice temperatures'). Borane (BH3, made from sodium borohydride) is a selective reagent and may work, as may the perfluoro analogue of STAB, which can probably be made using trifluoroacetic acid instead of acetic acid in Barium's procedure.
The only other method I can think of which may be capable is catalytic hydrogenation, using a carefully chosen catalyst. But to the best of my knowledge there are no literature references for this having been carried out, and you would have to do a good deal of experimenting before you found suitable conditions.
There are - without a doubt - many interesting compounds which are still not listed as scheduled drugs in the US. However, you must be aware these substances are exactly what the Controlled Substances Analogues Act of 1986 was designed to stamp out. You may still be prosecuted for making something which isn't yet controlled. I take it you've read the extensions and commentary to AET in PiHKAL?
As for recommendations, it depends on what you like. Are J (BDB) and Methyl-J (MBDB) controlled in the US? The latter is probably the closest you will come to a legal analogue of MDMA. Unfortunately I haven't tried either, as they are both controlled over here (as is everything else in PiHKAL). Maybe the methylenedioxyaminoindanes or tetralins will substitute for MDMA in humans.
There are still many of the 2C-X and DOX series which are not controlled in the US, for example 2C-N and DON. The BOX family is something I hope holds promise, as some novel BOX analogues are not controlled where I live. IAP is legal and moderately interesting; and Nichols' beautiful dragonfly creation is somewhere on my 'to-do' list. Here are some of their structures, which I hope helps you to visualise things a bit better:
Molecule:
MDMA and FLEA ("c1c(ccc2c1OCO2)CC(C)NC.c1c(ccc2c1OCO2)CC(C)N(C)O")
Molecule:
J (BDB) and Methyl-J (MBDB) ("c1c(ccc2c1OCO2)CC(CC)N.c1c(ccc2c1OCO2)CC(CC)NC")
Molecule:
Methylenedioxyaminoindane and methylenedioxyaminotetralin ("c1c3c(cc2c1OCO2)CC(C3)N.c1c3c(cc2c1OCO2)CCC(C3)N")
Molecule:
2C-N and DON ("c1(c(cc(c(c1)[N+]([O-])=O)OC)CCN)OC.c1(c(cc(c(c1)[N+]([O-])=O)OC)CC(N)C)OC ")
Molecule:
BO-something and Nichols' dragonfly ("c1(c(cc(cc1)OC)C(CN)OC)OC.c12c(c3c(cc1cco2)occ3)CC(N)C ")
For the last picture, the BOX analogue has yet to be named as BOH is already taken. It, along with the dragonfly molecule, can be substituted in
para- to the aminoethyl or propyl chain with a variety of atoms.
Related reading:AET
(
http://www.erowid.org/library/books_online/tihkal/tihkal11.shtml)
J (BDB)
(
http://www.erowid.org/library/books_online/pihkal/pihkal094.shtml)
Methyl-J (MBDB)
(
http://www.erowid.org/library/books_online/pihkal/pihkal128.shtml)
2C-N
(
http://www.erowid.org/library/books_online/pihkal/pihkal034.shtml)
DON
(
http://www.erowid.org/library/books_online/pihkal/pihkal070.shtml)
IAP
(
https://www.thevespiary.org/rhodium/Rhodium/chemistry/iap.html)
Methylenedioxyaminoindanes and tetralins
(
https://www.thevespiary.org/rhodium/Rhodium/pdf/nichols/nichols-benzofuran.indan.tetralin.mda-analogs.pdf)
Nichols' dragonfly
(
https://www.thevespiary.org/rhodium/Rhodium/pdf/nichols/nichols-dragonfly-2.pdf)
Good luck.
