1) djvu bigger than PDF.
As I understand it, djvu is "an image compression technology". PDF is a portable document format. PDF can store text as character strings and the viewer can render those strings using external or embedded fonts. PDF can include compressed bitmaps of scanned text but that isn't the only way it can store text.
If you take a PDF where the text is strings+fonts and convert it to djvu, maybe (under some circumstances) the strings have to be rasterised - thereby increasing the size of data.
If you can provide a small example PDF that exhibits the behaviour, people may be able to explain in more detail.
2) Lossless compression
There are many ways to make things smaller without losing information. some compression algorithms are more efficient than others for certain types of data.