This is because conceptually a TWebCheckBox in TMS WEB Core is modelled after a VCL TCheckBox and a checkbox in HTML is just a checkbox without caption. So, there needs to be a label HTML element also if you want a caption mapped on it.
The only thing we can do is make this smarter in the next version to allow either mapping on a single HTML INPUT element where the caption will NOT be linked but is assumed to be set in HTML or link to a SPAN element that has a first child HTML INPUT CHECKBOX element and a 2nd element a SPAN to hold the label.