[Genome] knownIsoforms

[Jeffrey Rosenfeld]

How is the knownisoforms table constructed? It seems that all 
overlapping genes are clustered together, but there are examples, such 
as at the very beginning of chromosome 1, where a transcript within a 
cluster becomes its own cluster.  When I run the following query on hg18:

select clusterID,name,chrom,strand,txStart,txEnd,cdsStart,cdsEnd  from 
knownIsoforms, knownGene where transcript = name;

These are the results I get:

| clusterID | name       | chrom | strand | txStart | txEnd  | cdsStart 
| cdsEnd |
+-----------+------------+-------+--------+---------+--------+----------+--------+
|         1 | uc001aaa.1 | chr1  | +      |    1736 |   4121 |     1736 
|   1736 |
|         2 | uc001aab.1 | chr1  | -      |    4558 |  14764 |     4558 
|   4558 |
|         2 | uc001aac.1 | chr1  | -      |    4558 |  19346 |     4558 
|   4558 |
|         2 | uc001aad.1 | chr1  | -      |    4558 |   7231 |     4558 
|   7173 |
|         2 | uc001aae.1 | chr1  | -      |    4558 |   9622 |     4558 
|   4558 |
|         2 | uc001aaf.1 | chr1  | -      |    4832 |  19672 |     4832 
|   4832 |
|         2 | uc001aag.1 | chr1  | -      |    5658 |   7231 |     5658 
|   5658 |
|         2 | uc001aah.1 | chr1  | -      |    6720 |  19346 |     6720 
|   6720 |
|         2 | uc001aai.1 | chr1  | -      |    6720 |   9622 |     6720 
|   6720 |
|         3 | uc001aaj.1 | chr1  | -      |    7777 |  19346 |     7777 
|  14749 |


Shouldn't cluster 3 be included as part of cluster 4?

Thank You,

Jeffrey Rosenfeld