It's very rare for  an Israeli geographic name to  be established in  anything  but  standard  romanization.   Why  is  this?   The
operative RI is 23.2 (p. 1), which states:

      For other names (i.e. those not in  the United States, Great  Britain,  Australia,  New  Zealand,  or  Canada),  base  the
      heading either on the form found in the work being cataloged  together  with  a consideration  of  the  form  found  in  a
      recently published gazetteer *or*  on  the  form provided by  the  U.S.  Board on Geographic Names  (BGN),  as
      necessary. This means that BGN *may* be consulted routinely or not,  as  efficiency, etc., in  the particular case warrants.
      It also means that all problems not resolved by the work in hand and a recently published gazetteer *must* be
      referred to BGN.

Specifics of  LC Hebraica Team practice regarding the establishment of geographic names are  given  in *Hebraica  Cataloging", p. 66-67, note on Gazetteers. We apply the option of consulting the BGN Gazetteer  of  Israel  (2nd  ed.,  1983)  for  all place-names in Israel.  (We also use  the 1992 Gazetteer of  the  West Bank  and Gaza Strip).

But how do we interpret the directive in the RI  to consider the  form found  in  the  work *together with* the  form provided by BGN?

In  practice  this  comparative  consideration  produces  an established form which is
(a)  in ALA-LC (also called "standard"  or "systematic") romanization,
(b)  as found  in  the  work being  cataloged,
(c)  *unless*  BGN  provides  a non-standard romanized  form and marks it "conventional."
The condition stated in (c) is  rare.

As  an example, suppose we want  to establish a  city called  "Hefah" in  the  work being cataloged (in  fact,  this particular
city has already been established).  We find  on  p.  62  of  the  Gazetteer the  entry "Hefa"  (H underlined-- this  is  BGN's  own  romanization, not the same as LC's by any means) with "Hebrew" in  brackets after it,  and following this after a semi-colon "Haifa"  with "conventional" in brackets.  In this case we  are authorized  to establish the  name  as "Haifa (Israel)" making refs from  the standard romanization ("Hefah," subscript dot  under  the  H)  of  what's in the work being cataloged, the BGN romanization ("Hefa,"   H underlined) and any other variants we're aware of.  If  BGN had   *not* provided a "conventional" form,  we'd  have established the  place as "Hefah (Israel)" (subscript dot under the  H)  in ALA-LC
romanization, and made refs from  the  BGN romanization and other   variants.

The address from which  the gazetteers with "names approved by  the  United States  Board on Geographic Names" can be ordered are as following:

     United States Geological Survey
     Branch of Information Services
     Box 25286
     Denver, CO 80225

          P.S. You  can call 1-800-USA-MAPS and perhaps get  more  information than this, like prices.

The  two  of interest to  us  are Gazetteer of  Israel, 2nd  ed.,  September 1983,  stock  number GAZGNISRAEL, and Gazetteer of  the  West   Bank   and   Gaza  Strip,  December  1992,   stock  number    GAZGNWESTBGAZA.
They  are publications of  the  Defense  Mapping  Agency.

---------------------------------------
  .
The web site  www.jewishgen.org/ShtetlSeeker  provides variant names and map coordinates, and even a
listing of towns in a stated geographical area.  It claims to provide BGN-approved names if they exist, but these will require      cross-checking.  The site won't substitute for consulting LC when you want to establish a shtetl, or update a heading to its post-Soviet-breakup form.  But it's a place to start and a source of variant romanizations.

Hebrew names of jurisdictions, except in the rare rare cases where  there's a "conventional name", are established in standard    romanization, which includes hyphenation and capitalization, etc. according to AACR2rev.  If the standard romanization normalizes to something different from the GeoNet form (as in this case), make a ref  from the GeoNet form.