We present 77 new granite whole-rock analyses from the Qattar and Gharib areas, Eastern Desert, Egypt. Both areas include a "normal" granite and either a hypersolvus (Gharib) or an almost plagioclase-free granite (Qattar) enriched in fluorite. According to earlier results, F influences element distribution in granitic melts forming complexes with specific elements as Nb, Ta, Ga, Hf, Th, Zn, Sn, whereas F excludes Ba and Sr. We use principal component analyses to split the granite into chemical groups allowing an unbiased study of the inter-group element distribution. This adds the heavy REEs and Y to the earlier lists of elements with an affinity for F. The light REEs show a decreasing affinity with decreasing atomic mass; fluorine separates Sm from Nd, whereas Zr follows La. Opposite to some, but in accordance with other earlier results, the ratio Nb/Ta is higher in the fluorite-enriched than in the other granite. Weak tetrad effects are present. Zircon in the hypersolvus granite is high in common lead. We suggest F to be instrumental for separating Pb 2+ from Pb 4+ . Two hypotheses may explain the occurrence of the two contrasting granites: they have either different sources, or they are co-magmatic, but the magma was split into two discrete types. We apply the second hypothesis as our working hypothesis. The liquidus has a gentler slope with pressure than the diapir requiring crystallisation to be most important in the lower part of the magma chamber. Our hypothesis suggests that globules of magma, enriched in volatile components, form during crystallisation due to slow diffusion rates in the crystallizing magma. Elements accompanying F are distributed into this magma batch, which has a lowered density and viscosity than the rest of the magma due to its increased contents of volatile components. A mushroom-formed diapir rises, forming the hypersolvus (or almost plagioclase-free) granite. Due to an edge effect, it is concentrated close to the wall of the magma chamber. The size and form of the outcropping granite depend on the intersection of the diapir with the erosion surface. Fluorine only makes it possible to follow the process. The model may be generalised to explain the diversification of non-F enriched granite, since the buoyancy of a magma batch several thousand m 3 in size has a much larger impact on the system than the small negative buoyancy of crystals or small crystal aggregates. A-type granite classified merely from its trace element content may form from separated F-enriched magma batches. This may be the reason for their high frequency in the Eastern Desert.