I'm seeing some odd behavior in Pipeline Pilot 9.1. I've attached a protocol to illustrate.

First, the PP handling of NOT queries, that is, that notH is inserted automatically even if it doesn't exist in the SMARTS string, is being performed as described by Accelrys in a previous discussion here.

However, there is still some strange behavior (see the attached protocol). Run the protocol, and then see.

1. If the query is represented in Kekule form, the fourth compound should satisfy the query, and doesn't. You see the same behavior whether or not the test compounds are in Kekule or aromatic representation.
2. If the query is represented in an aromatic form, the hits are appropriate according to the addition of an implicit notH. (again whether or not the test molecules are aromatic or Kekule)

By the way, if you go to the Daylight web page http://www.daylight.com/daycgi_tutorials/depictmatch.cgi , you can test these thing out quickly. Use the following molecules:

Oc1ccc[nH]1

[H]c1ccc(O)[nH]1

Cc1ccc(O)[nH]1

Oc1ccc(Cl)[nH]1

and then the following queriey:

[#8]-c1ccc(-[!#6])n1

Again, Accelrys is correct that notH is being inserted even if not specified in the query. (Note that in this web page, we cannot examine the effect of aromatic vs. kekule representation of the molecules, since everything is converted to aromatic automatically.)

Please note that Daylight DOES have a an explicit-H smarts mode in the DEPICTMATCH web page, in which case the queries behave in a way that Willem and I have requested.