ThunderSTORM (latest build from 07-27-16), latest Fiji (ImageJ 1.51f) under Java 8 (102) on OSX El Capitan (10.11.5).
I think there is a bug in the "Normalized Gaussian" reconstruction mode. When the "force" option for the lateral uncertainty is unchecked, the reconstruction is supposed to be made from the localization results "uncertainty_xy" value (on a per-localization basis). This is the case if the "Lateral Uncertainty" box contains a large value (larger than the typical uncertainties contained in the localizations results). However, if this box contains a small value (for example 1 nm), then it modifies the reconstruction even if the "force" box is unchecked. The resulting Gaussians aren't normalized anymore! I have the impression that it limits the plotting to 3*1 nm, but with the intensity of the larger gaussian (corresponding to the uncertainty_xy value found in the localization results).
Here is how to replicate:
Download the attached Test.zip folder and unzip it:
Test.zip
Open the test macro _TestMacro.ijm with ImageJ.
Edit the line 4 (run("Import"...) by inserting the path to the test localizations results file _TestTS.xls (tab-delimited text file) contained in the unzipped folder.
Set the pixel size in "Plugins>ThunderSTORM>Camera setup" to 160 nm for proper results.
Run the macro.
The test file has two rows of five localizations with uncertainties of 4, 8, 16, 32 and 64 nm. You should obtain two reconstructions: both are with "force" unchecked, but one with "1.0" nm in the Lateral Uncertainty box (corresponding to the attached _Test_TS(001)Rec.tif), the other with "100" nm (corresponding to the attached _Test_TS(100)Rec.tif). The correct reconstruction is "Test_TS(100)_Rec":
whereas "Test_TS(001)_Rec" is wrong:
As it is, having a small value in the "Lateral Uncertainty" box can generate completely wrong reconstructions, so I think it is a serious bug.
zitmen
commented
7 years ago 
ThunderSTORM (latest build from 07-27-16), latest Fiji (ImageJ 1.51f) under Java 8 (102) on OSX El Capitan (10.11.5).
I think there is a bug in the "Normalized Gaussian" reconstruction mode. When the "force" option for the lateral uncertainty is unchecked, the reconstruction is supposed to be made from the localization results "uncertainty_xy" value (on a per-localization basis). This is the case if the "Lateral Uncertainty" box contains a large value (larger than the typical uncertainties contained in the localizations results). However, if this box contains a small value (for example 1 nm), then it modifies the reconstruction even if the "force" box is unchecked. The resulting Gaussians aren't normalized anymore! I have the impression that it limits the plotting to 3*1 nm, but with the intensity of the larger gaussian (corresponding to the uncertainty_xy value found in the localization results).
Here is how to replicate:
The test file has two rows of five localizations with uncertainties of 4, 8, 16, 32 and 64 nm. You should obtain two reconstructions: both are with "force" unchecked, but one with "1.0" nm in the Lateral Uncertainty box (corresponding to the attached _Test_TS(001)Rec.tif), the other with "100" nm (corresponding to the attached _Test_TS(100)Rec.tif). The correct reconstruction is "Test_TS(100)_Rec":
whereas "Test_TS(001)_Rec" is wrong:
As it is, having a small value in the "Lateral Uncertainty" box can generate completely wrong reconstructions, so I think it is a serious bug.