creating tables with distinct tuples

I am trying to create one table from another in such a way that it
contains only a subset of distinct rows from the first.  I load data
into the first table, then 'insert into table select distinct ...' and
expected to find a bunch of distinct rows.  No such luck for my data
set (though it works with other test data sets)!  Am I missing
something obvious about how to do this?  An outline of what I'm doing
follows, but I've left out the data.  So far I cannot make a small
data set exhibit the problem.

Any help greatly appreciated!

Cheers,
Brook

---------------------------------------------------------------------------
drop table leaf_surveys;
create table leaf_surveys
(
 leaf_survey_id    int4    primary key        default nextval ('leaf_survey_id_sequence'),
 user        name                default current_user,
 time        datetime            default datetime (now ()),

 plot_id    int4    not null,        -- references plots.plot_id
 id        int    not null,
 plant_number    int,
 tag        int4,

 survey_date    text,
 color        char(1),
 status        int,
 leaf_number    int,
 stalk_number    int,

 unique (plot_id, id, survey_date)
);
---------------------------------------------------------------------------
insert into leaf_surveys (plot_id, id, plant_number, tag, survey_date, color, status, leaf_number, stalk_number)
     values ('903', '1', '', '82', '05/24/95', '', '2', '7', '1');
insert into leaf_surveys (plot_id, id, plant_number, tag, survey_date, color, status, leaf_number, stalk_number)
     values ('903', '2', '', '97', '05/24/95', '', '2', '7', '');
insert into leaf_surveys (plot_id, id, plant_number, tag, survey_date, color, status, leaf_number, stalk_number)
     values ('903', '4', '', '148', '05/24/95', '', '2', '12', '0');
-- lots more omitted
---------------------------------------------------------------------------
drop table plants;
create table plants
(
 plant_id    int4    primary key        default nextval ('plant_id_sequence'),
 user        name                default current_user,
 time        datetime            default datetime (now ()),

 plot_id    int4    not null,        -- references plots.plot_id
 id        int4,
 tag        int4
-- unique (plot_id, id, tag)
);
---------------------------------------------------------------------------
select distinct plot_id, id, tag from leaf_surveys; -- yields a list of distinct rows
insert into plants (plot_id, id, tag) select distinct plot_id, id, tag from leaf_surveys;
select * from plants; -- yields a list with duplicate rows