Interesting radiation at dusk

Some odd artifacts in Rsw in the transition between when direct sun is blocked by the mountains, but the sun hasn't set. 

First, from several days ago, I see the amount of incoming (diffuse) radiation higher at, say 15:40 than it was at 15:00.  Perhaps the amount of cirrus cloud coverage changed during this period?

Second, today I see Rsw.out.9m higher than the incomings at 15:00.  Usually, I'd blame this on snow covering the incoming sensors, but Rsw.in.9m is within a CVF4 housing that <should> be keeping it clean.  Also, Rsw.in from the NR01 tracks Rsw.in.9m quite well, even though it has no housing and a different construction.  Thus, I'm inclined to think that this is a real reflection somehow.

P.S. Note the large change in albedo between these two days, due to the recent snow.

P.P.S. The huge dip in Rsw.in.uw at 09:30 is due to the earlier-mentioned shading by the uw tower.