the total number of active contributors with IQ above 132 or something like that might be better

Actually, a similar concern was brought up in response to my IQ Accuracy comment and Vaniver discovered that the average IQs of the active members and lurkers was almost exactly the same:

165 out of 549 responses without reported positive karma (30%) self-reported an IQ score; the average response was 138.44.

181 out of 518 responses with reported positive karma (34%) self-reported an IQ score; the average response was 138.25.

We could separate the lurkers from the active members and do the analysis again, but I'm not sure it would be worth the effort as it looks to me like active members and lurkers are giving similar answers. If you'd like to do that, I'd certainly be interested in any surprises you uncover, but I don't expect it to be worthwhile enough to do it myself.

I'd guess that the decline in average IQ is mostly due to lower-IQ people arriving rather than to higher-IQ people leaving (EDIT: applying the intraocular trauma test to this graph appears to confirm that)

The sample set for the highest IQ groups is, of course, rather small, but what's been happening with the highest IQ groups is not encouraging. The specific graph in question (although I very much doubt that Gwern would intend to make that graph misleading in any way) is just not designed to clearly illustrate that particular aspect of the results visually.

Here are a few things you wouldn't guess without looking at the numbers:

Exceptionally gifted people used to be 18% of the IQ respondents. Now they are 6%.

The total number of highly and exceptionally gifted respondents decreased in 2012, while normal and moderately gifted respondents increased.

I did some analysis here