Enumerating still life objects in Game of Life

Back in the mid-seventies, Peter Raynham wrote the first still-life enumeration program, which was used to extend the list of known still-lifes to include all still-lifes up to 14 bits.
In 1989, Mark Niemiec wrote a similar program, and enumerated all still-lifes up to 18 bits. (It generated up to 20 bits, but didn't have enough disk space to save the intermediate results.)
Different versions of this program have, over time, generated the same list of still-lifes, with possible differences in how to interpret configurations consisting of two or more islands.
The program only generates adjacent islands, if at least one of the islands requires the existence of the other. This eliminates trivial non-objects such as:

  aa.bb
  aa.bb

This is done by treating objects as directed graphs of islands; for example,

  aa.bb
  .a.bb
  .a...
  aa...

generates the graph A->B; if the table (A) is generated first, the block (B) is naturally tried as a stabilizer for it; however, if the block is generated first, there is no need to add the table. This causes problems in objects of the form A->B<-C, which are never found. These types of objects must be added manually (starting at 16 bits and up):

  aa.bb.cc
  .a.bb.c.
  .a....c.
  aa....cc

The program also erroneously reports objects of the following type, in which islands are generated to unnecessarily stabilize neighbourhoods of 4 or more, which are already stable. These types of extra non-objects must be removed manually:

  aa.a..  aa.b...  aa.a.aa
  a.aa..  .a.bbb.  a.aaa.a
  ....aa  a.....b  .......
  .bb.a.  .aaa.b.  .aa.bb.
  .bb..a  ...a.bb  .aa.bb.
  ....aa

(Perhaps a better approach would be to generate all still-lifes including pseudo-objects such as the block-on-block above, and externally filter out all with mutually exclusive sets of islands; this would solve both of the above problems.)
My statistics are as follows:

Bits:      4  5  6  7  8   9  10  11  12  13  14   15   16   17    18
Gross:     2  1  5  4  9  10  25  46 121 240 619 1353 3286 7778 19027
+Missing:  0  0  0  0  0   0   0   0   0   0   0    0    2    0    25
-Extras:   0  0  0  0  0   0   0   0   0   0   0    0    2    3    10

Net:       2  1  5  4  9  10  25  46 121 240 619 1353 3286 7775 19042

(Mark has recently re-run the 4-18's to verify the counts.)
Several years later, he re-wrote the still-life enumerator for more efficient tree-pruning (the CPU time is O(2.4^n) as opposed to O(2.9^n) for the above-mentioned program, and O(4^n) for Peter Raynham's, and since the number of still-lifes also is about O(2.4^n), he believes this to be optimal). This program also searches for still-lifes in 3-dimensional space (x, y, time), effectively finding oscillators and spaceships. Unfortunately, for periods >2, the increase in CPU time for each additional cell makes the program unusable for finding anything other than still-lifes and period-2 oscillators.

In a letter dated from 28.2.1997 M. D. Niemiec gave corrected/further values based on more sophisticate computations and an intensive discussion with H. K"onig.

Bits:    4  5  6  7  8   9  10  11  12  13  14   15   16   17    18     19     20     21
Count:   2  1  5  4  9  10  25  46 121 240 619 1353 3286 7773 19044  45759 112243 273188
Pseudo:              1   1   7  16  55 110 279  620 1645 4067 10843  27250  70637 179011

Mark D. Niemiec